2023-12-31 13:55:22 +03:00
// NOTE: GENERATED by github.com/mjl-/sherpats, DO NOT MODIFY
namespace api {
// Domain is a domain name, with one or more labels, with at least an ASCII
// representation, and for IDNA non-ASCII domains a unicode representation.
// The ASCII string must be used for DNS lookups. The strings do not have a
// trailing dot. When using with StrictResolver, add the trailing dot.
export interface Domain {
ASCII : string // A non-unicode domain, e.g. with A-labels (xn--...) or NR-LDH (non-reserved letters/digits/hyphens) labels. Always in lower case. No trailing dot.
Unicode : string // Name as U-labels. Empty if this is an ASCII-only domain. No trailing dot.
}
export interface Destination {
Mailbox : string
Rulesets? : Ruleset [ ] | null
FullName : string
}
export interface Ruleset {
SMTPMailFromRegexp : string
VerifiedDomain : string
HeadersRegexp ? : { [ key : string ] : string }
IsForward : boolean // todo: once we implement ARC, we can use dkim domains that we cannot verify but that the arc-verified forwarding mail server was able to verify.
ListAllowDomain : string
AcceptRejectsToMailbox : string
Mailbox : string
VerifiedDNSDomain : Domain
ListAllowDNSDomain : Domain
}
// ImportProgress is returned after uploading a file to import.
export interface ImportProgress {
Token : string // For fetching progress, or cancelling an import.
}
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
export type CSRFToken = string
2023-12-31 13:55:22 +03:00
export const structTypes : { [ typename : string ] : boolean } = { "Destination" : true , "Domain" : true , "ImportProgress" : true , "Ruleset" : true }
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
export const stringsTypes : { [ typename : string ] : boolean } = { "CSRFToken" : true }
2023-12-31 13:55:22 +03:00
export const intsTypes : { [ typename : string ] : boolean } = { }
export const types : TypenameMap = {
"Domain" : { "Name" : "Domain" , "Docs" : "" , "Fields" : [ { "Name" : "ASCII" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "Unicode" , "Docs" : "" , "Typewords" : [ "string" ] } ] } ,
"Destination" : { "Name" : "Destination" , "Docs" : "" , "Fields" : [ { "Name" : "Mailbox" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "Rulesets" , "Docs" : "" , "Typewords" : [ "[]" , "Ruleset" ] } , { "Name" : "FullName" , "Docs" : "" , "Typewords" : [ "string" ] } ] } ,
"Ruleset" : { "Name" : "Ruleset" , "Docs" : "" , "Fields" : [ { "Name" : "SMTPMailFromRegexp" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "VerifiedDomain" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "HeadersRegexp" , "Docs" : "" , "Typewords" : [ "{}" , "string" ] } , { "Name" : "IsForward" , "Docs" : "" , "Typewords" : [ "bool" ] } , { "Name" : "ListAllowDomain" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "AcceptRejectsToMailbox" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "Mailbox" , "Docs" : "" , "Typewords" : [ "string" ] } , { "Name" : "VerifiedDNSDomain" , "Docs" : "" , "Typewords" : [ "Domain" ] } , { "Name" : "ListAllowDNSDomain" , "Docs" : "" , "Typewords" : [ "Domain" ] } ] } ,
"ImportProgress" : { "Name" : "ImportProgress" , "Docs" : "" , "Fields" : [ { "Name" : "Token" , "Docs" : "" , "Typewords" : [ "string" ] } ] } ,
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
"CSRFToken" : { "Name" : "CSRFToken" , "Docs" : "" , "Values" : null } ,
2023-12-31 13:55:22 +03:00
}
export const parser = {
Domain : ( v : any ) = > parse ( "Domain" , v ) as Domain ,
Destination : ( v : any ) = > parse ( "Destination" , v ) as Destination ,
Ruleset : ( v : any ) = > parse ( "Ruleset" , v ) as Ruleset ,
ImportProgress : ( v : any ) = > parse ( "ImportProgress" , v ) as ImportProgress ,
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
CSRFToken : ( v : any ) = > parse ( "CSRFToken" , v ) as CSRFToken ,
2023-12-31 13:55:22 +03:00
}
// Account exports web API functions for the account web interface. All its
// methods are exported under api/. Function calls require valid HTTP
// Authentication credentials of a user.
let defaultOptions : ClientOptions = { slicesNullable : true , mapsNullable : true , nullableOptional : true }
export class Client {
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
private baseURL : string
public authState : AuthState
public options : ClientOptions
constructor ( ) {
this . authState = { }
this . options = { . . . defaultOptions }
this . baseURL = this . options . baseURL || defaultBaseURL
}
withAuthToken ( token : string ) : Client {
const c = new Client ( )
c . authState . token = token
c . options = this . options
return c
2023-12-31 13:55:22 +03:00
}
withOptions ( options : ClientOptions ) : Client {
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
const c = new Client ( )
c . authState = this . authState
c . options = { . . . this . options , . . . options }
return c
}
// LoginPrep returns a login token, and also sets it as cookie. Both must be
// present in the call to Login.
async LoginPrep ( ) : Promise < string > {
const fn : string = "LoginPrep"
const paramTypes : string [ ] [ ] = [ ]
const returnTypes : string [ ] [ ] = [ [ "string" ] ]
const params : any [ ] = [ ]
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as string
}
// Login returns a session token for the credentials, or fails with error code
// "user:badLogin". Call LoginPrep to get a loginToken.
async Login ( loginToken : string , username : string , password : string ) : Promise < CSRFToken > {
const fn : string = "Login"
const paramTypes : string [ ] [ ] = [ [ "string" ] , [ "string" ] , [ "string" ] ]
const returnTypes : string [ ] [ ] = [ [ "CSRFToken" ] ]
const params : any [ ] = [ loginToken , username , password ]
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as CSRFToken
}
// Logout invalidates the session token.
async Logout ( ) : Promise < void > {
const fn : string = "Logout"
const paramTypes : string [ ] [ ] = [ ]
const returnTypes : string [ ] [ ] = [ ]
const params : any [ ] = [ ]
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as void
2023-12-31 13:55:22 +03:00
}
// SetPassword saves a new password for the account, invalidating the previous password.
// Sessions are not interrupted, and will keep working. New login attempts must use the new password.
// Password must be at least 8 characters.
async SetPassword ( password : string ) : Promise < void > {
const fn : string = "SetPassword"
const paramTypes : string [ ] [ ] = [ [ "string" ] ]
const returnTypes : string [ ] [ ] = [ ]
const params : any [ ] = [ password ]
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as void
2023-12-31 13:55:22 +03:00
}
// Account returns information about the account: full name, the default domain,
// and the destinations (keys are email addresses, or localparts to the default
// domain). todo: replace with a function that returns the whole account, when
// sherpadoc understands unnamed struct fields.
async Account ( ) : Promise < [ string , Domain , { [ key : string ] : Destination } ] > {
const fn : string = "Account"
const paramTypes : string [ ] [ ] = [ ]
const returnTypes : string [ ] [ ] = [ [ "string" ] , [ "Domain" ] , [ "{}" , "Destination" ] ]
const params : any [ ] = [ ]
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as [ string , Domain , { [ key : string ] : Destination } ]
2023-12-31 13:55:22 +03:00
}
async AccountSaveFullName ( fullName : string ) : Promise < void > {
const fn : string = "AccountSaveFullName"
const paramTypes : string [ ] [ ] = [ [ "string" ] ]
const returnTypes : string [ ] [ ] = [ ]
const params : any [ ] = [ fullName ]
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as void
2023-12-31 13:55:22 +03:00
}
// DestinationSave updates a destination.
// OldDest is compared against the current destination. If it does not match, an
// error is returned. Otherwise newDest is saved and the configuration reloaded.
async DestinationSave ( destName : string , oldDest : Destination , newDest : Destination ) : Promise < void > {
const fn : string = "DestinationSave"
const paramTypes : string [ ] [ ] = [ [ "string" ] , [ "Destination" ] , [ "Destination" ] ]
const returnTypes : string [ ] [ ] = [ ]
const params : any [ ] = [ destName , oldDest , newDest ]
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as void
2023-12-31 13:55:22 +03:00
}
// ImportAbort aborts an import that is in progress. If the import exists and isn't
// finished, no changes will have been made by the import.
async ImportAbort ( importToken : string ) : Promise < void > {
const fn : string = "ImportAbort"
const paramTypes : string [ ] [ ] = [ [ "string" ] ]
const returnTypes : string [ ] [ ] = [ ]
const params : any [ ] = [ importToken ]
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as void
2023-12-31 13:55:22 +03:00
}
// Types exposes types not used in API method signatures, such as the import form upload.
async Types ( ) : Promise < ImportProgress > {
const fn : string = "Types"
const paramTypes : string [ ] [ ] = [ ]
const returnTypes : string [ ] [ ] = [ [ "ImportProgress" ] ]
const params : any [ ] = [ ]
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
return await _sherpaCall ( this . baseURL , this . authState , { . . . this . options } , paramTypes , returnTypes , fn , params ) as ImportProgress
2023-12-31 13:55:22 +03:00
}
}
export const defaultBaseURL = ( function ( ) {
let p = location . pathname
if ( p && p [ p . length - 1 ] !== '/' ) {
let l = location . pathname . split ( '/' )
l = l . slice ( 0 , l . length - 1 )
p = '/' + l . join ( '/' ) + '/'
}
return location . protocol + '//' + location . host + p + 'api/'
} ) ( )
// NOTE: code below is shared between github.com/mjl-/sherpaweb and github.com/mjl-/sherpats.
// KEEP IN SYNC.
export const supportedSherpaVersion = 1
export interface Section {
Name : string
Docs : string
Functions : Function [ ]
Sections : Section [ ]
Structs : Struct [ ]
Ints : Ints [ ]
Strings : Strings [ ]
Version : string // only for top-level section
SherpaVersion : number // only for top-level section
SherpadocVersion : number // only for top-level section
}
export interface Function {
Name : string
Docs : string
Params : Arg [ ]
Returns : Arg [ ]
}
export interface Arg {
Name : string
Typewords : string [ ]
}
export interface Struct {
Name : string
Docs : string
Fields : Field [ ]
}
export interface Field {
Name : string
Docs : string
Typewords : string [ ]
}
export interface Ints {
Name : string
Docs : string
Values : {
Name : string
Value : number
Docs : string
} [ ] | null
}
export interface Strings {
Name : string
Docs : string
Values : {
Name : string
Value : string
Docs : string
} [ ] | null
}
export type NamedType = Struct | Strings | Ints
export type TypenameMap = { [ k : string ] : NamedType }
// verifyArg typechecks "v" against "typewords", returning a new (possibly modified) value for JSON-encoding.
// toJS indicate if the data is coming into JS. If so, timestamps are turned into JS Dates. Otherwise, JS Dates are turned into strings.
// allowUnknownKeys configures whether unknown keys in structs are allowed.
// types are the named types of the API.
export const verifyArg = ( path : string , v : any , typewords : string [ ] , toJS : boolean , allowUnknownKeys : boolean , types : TypenameMap , opts : ClientOptions ) : any = > {
return new verifier ( types , toJS , allowUnknownKeys , opts ) . verify ( path , v , typewords )
}
export const parse = ( name : string , v : any ) : any = > verifyArg ( name , v , [ name ] , true , false , types , defaultOptions )
class verifier {
constructor ( private types : TypenameMap , private toJS : boolean , private allowUnknownKeys : boolean , private opts : ClientOptions ) {
}
verify ( path : string , v : any , typewords : string [ ] ) : any {
typewords = typewords . slice ( 0 )
const ww = typewords . shift ( )
const error = ( msg : string ) = > {
if ( path != '' ) {
msg = path + ': ' + msg
}
throw new Error ( msg )
}
if ( typeof ww !== 'string' ) {
error ( 'bad typewords' )
return // should not be necessary, typescript doesn't see error always throws an exception?
}
const w : string = ww
const ensure = ( ok : boolean , expect : string ) : any = > {
if ( ! ok ) {
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
error ( 'got ' + JSON . stringify ( v ) + ', expected ' + expect )
2023-12-31 13:55:22 +03:00
}
return v
}
switch ( w ) {
case 'nullable' :
if ( v === null || v === undefined && this . opts . nullableOptional ) {
return v
}
return this . verify ( path , v , typewords )
case '[]' :
if ( v === null && this . opts . slicesNullable || v === undefined && this . opts . slicesNullable && this . opts . nullableOptional ) {
return v
}
ensure ( Array . isArray ( v ) , "array" )
return v . map ( ( e : any , i : number ) = > this . verify ( path + '[' + i + ']' , e , typewords ) )
case '{}' :
if ( v === null && this . opts . mapsNullable || v === undefined && this . opts . mapsNullable && this . opts . nullableOptional ) {
return v
}
ensure ( v !== null || typeof v === 'object' , "object" )
const r : any = { }
for ( const k in v ) {
r [ k ] = this . verify ( path + '.' + k , v [ k ] , typewords )
}
return r
}
ensure ( typewords . length == 0 , "empty typewords" )
const t = typeof v
switch ( w ) {
case 'any' :
return v
case 'bool' :
ensure ( t === 'boolean' , 'bool' )
return v
case 'int8' :
case 'uint8' :
case 'int16' :
case 'uint16' :
case 'int32' :
case 'uint32' :
case 'int64' :
case 'uint64' :
ensure ( t === 'number' && Number . isInteger ( v ) , 'integer' )
return v
case 'float32' :
case 'float64' :
ensure ( t === 'number' , 'float' )
return v
case 'int64s' :
case 'uint64s' :
ensure ( t === 'number' && Number . isInteger ( v ) || t === 'string' , 'integer fitting in float without precision loss, or string' )
return '' + v
case 'string' :
ensure ( t === 'string' , 'string' )
return v
case 'timestamp' :
if ( this . toJS ) {
ensure ( t === 'string' , 'string, with timestamp' )
const d = new Date ( v )
if ( d instanceof Date && ! isNaN ( d . getTime ( ) ) ) {
return d
}
error ( 'invalid date ' + v )
} else {
ensure ( t === 'object' && v !== null , 'non-null object' )
ensure ( v . __proto__ === Date . prototype , 'Date' )
return v . toISOString ( )
}
}
// We're left with named types.
const nt = this . types [ w ]
if ( ! nt ) {
error ( 'unknown type ' + w )
}
if ( v === null ) {
error ( 'bad value ' + v + ' for named type ' + w )
}
if ( structTypes [ nt . Name ] ) {
const t = nt as Struct
if ( typeof v !== 'object' ) {
error ( 'bad value ' + v + ' for struct ' + w )
}
const r : any = { }
for ( const f of t . Fields ) {
r [ f . Name ] = this . verify ( path + '.' + f . Name , v [ f . Name ] , f . Typewords )
}
// If going to JSON also verify no unknown fields are present.
if ( ! this . allowUnknownKeys ) {
const known : { [ key : string ] : boolean } = { }
for ( const f of t . Fields ) {
known [ f . Name ] = true
}
Object . keys ( v ) . forEach ( ( k ) = > {
if ( ! known [ k ] ) {
error ( 'unknown key ' + k + ' for struct ' + w )
}
} )
}
return r
} else if ( stringsTypes [ nt . Name ] ) {
const t = nt as Strings
if ( typeof v !== 'string' ) {
error ( 'mistyped value ' + v + ' for named strings ' + t . Name )
}
if ( ! t . Values || t . Values . length === 0 ) {
return v
}
for ( const sv of t . Values ) {
if ( sv . Value === v ) {
return v
}
}
error ( 'unknkown value ' + v + ' for named strings ' + t . Name )
} else if ( intsTypes [ nt . Name ] ) {
const t = nt as Ints
if ( typeof v !== 'number' || ! Number . isInteger ( v ) ) {
error ( 'mistyped value ' + v + ' for named ints ' + t . Name )
}
if ( ! t . Values || t . Values . length === 0 ) {
return v
}
for ( const sv of t . Values ) {
if ( sv . Value === v ) {
return v
}
}
error ( 'unknkown value ' + v + ' for named ints ' + t . Name )
} else {
throw new Error ( 'unexpected named type ' + nt )
}
}
}
export interface ClientOptions {
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
baseURL? : string
2023-12-31 13:55:22 +03:00
aborter ? : { abort ? : ( ) = > void }
timeoutMsec? : number
skipParamCheck? : boolean
skipReturnCheck? : boolean
slicesNullable? : boolean
mapsNullable? : boolean
nullableOptional? : boolean
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
csrfHeader? : string
login ? : ( reason : string ) = > Promise < string >
2023-12-31 13:55:22 +03:00
}
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
export interface AuthState {
token? : string // For csrf request header.
loginPromise? : Promise < void > // To let multiple API calls wait for a single login attempt, not each opening a login popup.
}
const _sherpaCall = async ( baseURL : string , authState : AuthState , options : ClientOptions , paramTypes : string [ ] [ ] , returnTypes : string [ ] [ ] , name : string , params : any [ ] ) : Promise < any > = > {
2023-12-31 13:55:22 +03:00
if ( ! options . skipParamCheck ) {
if ( params . length !== paramTypes . length ) {
return Promise . reject ( { message : 'wrong number of parameters in sherpa call, saw ' + params . length + ' != expected ' + paramTypes . length } )
}
params = params . map ( ( v : any , index : number ) = > verifyArg ( 'params[' + index + ']' , v , paramTypes [ index ] , false , false , types , options ) )
}
const simulate = async ( json : string ) = > {
const config = JSON . parse ( json || 'null' ) || { }
const waitMinMsec = config . waitMinMsec || 0
const waitMaxMsec = config . waitMaxMsec || 0
const wait = Math . random ( ) * ( waitMaxMsec - waitMinMsec )
const failRate = config . failRate || 0
return new Promise < void > ( ( resolve , reject ) = > {
if ( options . aborter ) {
options . aborter . abort = ( ) = > {
reject ( { message : 'call to ' + name + ' aborted by user' , code : 'sherpa:aborted' } )
reject = resolve = ( ) = > { }
}
}
setTimeout ( ( ) = > {
const r = Math . random ( )
if ( r < failRate ) {
reject ( { message : 'injected failure on ' + name , code : 'server:injected' } )
} else {
resolve ( )
}
reject = resolve = ( ) = > { }
} , waitMinMsec + wait )
} )
}
// Only simulate when there is a debug string. Otherwise it would always interfere
// with setting options.aborter.
let json : string = ''
try {
json = window . localStorage . getItem ( 'sherpats-debug' ) || ''
} catch ( err ) { }
if ( json ) {
await simulate ( json )
}
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
const fn = ( resolve : ( v : any ) = > void , reject : ( v : any ) = > void ) = > {
let resolve1 = ( v : any ) = > {
2023-12-31 13:55:22 +03:00
resolve ( v )
resolve1 = ( ) = > { }
reject1 = ( ) = > { }
}
let reject1 = ( v : { code : string , message : string } ) = > {
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
if ( ( v . code === 'user:noAuth' || v . code === 'user:badAuth' ) && options . login ) {
const login = options . login
if ( ! authState . loginPromise ) {
authState . loginPromise = new Promise ( ( aresolve , areject ) = > {
login ( v . code === 'user:badAuth' ? ( v . message || '' ) : '' )
. then ( ( token ) = > {
authState . token = token
authState . loginPromise = undefined
aresolve ( )
} , ( err : any ) = > {
authState . loginPromise = undefined
areject ( err )
} )
} )
}
authState . loginPromise
. then ( ( ) = > {
fn ( resolve , reject )
} , ( err : any ) = > {
reject ( err )
} )
return
}
2023-12-31 13:55:22 +03:00
reject ( v )
resolve1 = ( ) = > { }
reject1 = ( ) = > { }
}
const url = baseURL + name
const req = new window . XMLHttpRequest ( )
if ( options . aborter ) {
options . aborter . abort = ( ) = > {
req . abort ( )
reject1 ( { code : 'sherpa:aborted' , message : 'request aborted' } )
}
}
req . open ( 'POST' , url , true )
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
if ( options . csrfHeader && authState . token ) {
req . setRequestHeader ( options . csrfHeader , authState . token )
}
2023-12-31 13:55:22 +03:00
if ( options . timeoutMsec ) {
req . timeout = options . timeoutMsec
}
req . onload = ( ) = > {
if ( req . status !== 200 ) {
if ( req . status === 404 ) {
reject1 ( { code : 'sherpa:badFunction' , message : 'function does not exist' } )
} else {
reject1 ( { code : 'sherpa:http' , message : 'error calling function, HTTP status: ' + req . status } )
}
return
}
let resp : any
try {
resp = JSON . parse ( req . responseText )
} catch ( err ) {
reject1 ( { code : 'sherpa:badResponse' , message : 'bad JSON from server' } )
return
}
if ( resp && resp . error ) {
const err = resp . error
reject1 ( { code : err.code , message : err.message } )
return
} else if ( ! resp || ! resp . hasOwnProperty ( 'result' ) ) {
reject1 ( { code : 'sherpa:badResponse' , message : "invalid sherpa response object, missing 'result'" } )
return
}
if ( options . skipReturnCheck ) {
resolve1 ( resp . result )
return
}
let result = resp . result
try {
if ( returnTypes . length === 0 ) {
if ( result ) {
throw new Error ( 'function ' + name + ' returned a value while prototype says it returns "void"' )
}
} else if ( returnTypes . length === 1 ) {
result = verifyArg ( 'result' , result , returnTypes [ 0 ] , true , true , types , options )
} else {
if ( result . length != returnTypes . length ) {
throw new Error ( 'wrong number of values returned by ' + name + ', saw ' + result . length + ' != expected ' + returnTypes . length )
}
result = result . map ( ( v : any , index : number ) = > verifyArg ( 'result[' + index + ']' , v , returnTypes [ index ] , true , true , types , options ) )
}
} catch ( err ) {
let errmsg = 'bad types'
if ( err instanceof Error ) {
errmsg = err . message
}
reject1 ( { code : 'sherpa:badTypes' , message : errmsg } )
}
resolve1 ( result )
}
req . onerror = ( ) = > {
reject1 ( { code : 'sherpa:connection' , message : 'connection failed' } )
}
req . ontimeout = ( ) = > {
reject1 ( { code : 'sherpa:timeout' , message : 'request timeout' } )
}
req . setRequestHeader ( 'Content-Type' , 'application/json' )
try {
req . send ( JSON . stringify ( { params : params } ) )
} catch ( err ) {
reject1 ( { code : 'sherpa:badData' , message : 'cannot marshal to JSON' } )
}
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:
there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.
a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.
another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.
our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.
api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).
in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions. for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.
webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.
for issue #58
2024-01-04 15:10:48 +03:00
}
return await new Promise ( fn )
2023-12-31 13:55:22 +03:00
}
}