Re: RFC: AbiTalk communication protocol suggestion

From: <msevior_at_physics.unimelb.edu.au>
Date: Wed May 18 2005 - 09:38:30 CEST

>
>
> I have got really quite excited about this collaborative stuff, and
> since the proposal did not suggest the transport mechanism, here comes a
> suggestion for a basic protocol for communication between two AbiWord
> processes. Please note (a) this is only a top level protocol, which
> could be implemented on top of dbus or just a straigh TCP/IP, or some
> other protocol; (b) I might be up to no good, and it almost certainly
> not complete, hence RFC.
>
> The proposed protocol is xml-based, so that we would use an xml listener
> for parsing the messages as we do with AW importer; it means that our
> change records and similar items would be easy to translate into the
> protocol messages and vice versa.
>
> A message passed between two processes has the following format
> <abitalk version="1.0" [uid="xxxxx"] [token="xxxxx"]>
> <%cmd%>[content]</%cmd%>[<%cmd%>[content]</%cmd%>]
> </abitalk>
>
> version: identifies the protocol version
>
> uid: the uuid of the document which is the subject of conversation;
> if not provided, the commands to follow will carry operations
> only on the application, not documents.
>
> token: authorisation token. The token is a standard uuid with time
> stamp ajusted to indicate when the authoristation token
> expires. Tokens are connected to a specific document uid.
> If no token is provided the only legitimate commands to follow
> are <token> and <getdoc> (see below).
>
> Single message can contain one or more commands:
>
> %cmd%: specific command name, which can be followed by set of
> attributes
>
> [content]: content specific to given command (optional)
>
>
> Predefined commands
> -------------------
> The following commands are predefined by the protocol:
>
> <getdoc name="xxxx"/>
> Request for the entire document identified by name
>
> <document name="xxxx">[doc]</document>
> Sent in response to <getdoc> request. The content of this element
> is raw document as contained in the document file and should be
> passed directly to the AW importer chain.
>
> <token [uid="xxxx"]/>
> uid: the token in standard uuid format. The token is specific to
> the document uid, and has a limited lifespan; the expiry time is
> indicated by the uid time stamp.
>
> If no uid is specified, the command is interpreted as request for
> authorisation token.
>
> If uid is specified, this command is a response to request for
> authorisation token; uid is the token.
>
> <release uid="xxxx"/>
> uid: authorisation token.
>
> This command is sent to release the token; as soon soon as this
> command send, the token becomes invalid. This command is required
> for the entire transation to be valid (see below notes on token
> handling).
>
> <error id="errno">[Verbose msg]</error>
> id: UT_AbiTalk::Result
> Verbose msg: optional verbose, human readable, error message.
>
>
> All other commands are handled by command handlers registered via
> UT_AbiTalk::registerHandler(); their number and meaning has not direct
> impact on the protocol itself (i.e., the protocol can be easily expanded
> as new features are added).
>
> Token handling
> --------------
> A and B are communicating Abi processes. If A wishes to send commands to
> B, A requests an authorisitation token. B issues the token and until
> either the token expires or is released by A, B does not carry any other
> operations on the document/process than those that are requested by A
> using the authorisation token. At the same time, while A is in
> possession of the token, it will turn down any request for authorisation
> tokens from other parties, including B.
>
> When finished with the whole command sequence, A has to release the
> token by sending <release> command.
>
> B keeps an eye on the life of the issued token; if the token expires
> before the <release> cmd is received for the token, B reverses all
> operation carried on the document by A using that token and dispatches
> <error>Token expired</error> cmd to A. In response to the error A has to
> request a new token and resend all commands.
>
> Example command
> ---------------
> A sample command (after the intial hanshake and token retrieval) could
> look like:
>
> <abitalk version="1.0" uuui="xxxxxx" token="yyyyyy">
> <insertspan start=5 end=7 props="font-size:12pt">my span</insertspan>
> </abitalk>
>
> On receipt of the message, the AW process would translate it to an
> insert Span change record and pass it to the document identified by the
> uuid.

Hi Tomas,
         It's fantastic to have your help with this. I fully agree that we
want to roll your revisions, document tracking and even document
merging into this.

It would make the system extremely powerful.

Rather than time limitting the tokens I think we should let them die after
all the AbiWord's in the network have acknowledged receipt of them back to
the "central" server.

I also think we should only lock out changes that affect the same region
of the document as a currently valid token.

Changes outside this region can be handelled via the simple shuffling of
insertion points I talked about in linked document.

This means that we do not need to lock everybody in the network on every
keystroke. As long as the changes are made in different regions of the
document the position shuffling code should work.

Cheers

Martin

>
> Enough for one night.
>
> Tomas
>
>
>
>
Received on Wed May 18 09:38:58 2005

This archive was generated by hypermail 2.1.8 : Wed May 18 2005 - 09:38:58 CEST