RE: [Forces-protocol] RE: GET/SET in one msg ?

["Joel M. Halpern" <[e-mail]>]

I agree that rules about protocol combination are NOT for the protocol to 
enforce.  We would write them as ~the sender should not do X and Y, and 
receiver behavior if it receives such is indeterminant~.

I would agree that there should be no requirement on the order of 
application of operations in a message, although there probably is an 
atomicity requirement of the form ~either all of them will be done, or none 
of them will be done~.
Another way of putting this is that the protocol structures themselves 
should not mandate the presence of an order requirement, even if we as a 
working group were to decide that we wanted such ordering to be enforced.

Yours,
Joel

At 08:39 PM 10/15/2004 -0700, Deleganes, Ellen M wrote:
>The thing that is appealing about the proposal is not so much being able
>to combine operations as the notion of being able to define new ones
>without having to define new messages.
>
>I presume that the "rules" (e.g. never combine operations X and Y) are
>not something the protocol is expected to enforce. Otherwise, I think it
>would add unnecessary complication to the protocol processing.
>
>It also sounded like Joel was suggesting that if operations happened to
>be combined in a single message, the order shown in the protocol does
>not imply the order in which the implementation has to execute them. Is
>that right?
>
>Regards,
>Ellen
>
>
>-----Original Message-----
>From: Khosravi, Hormuzd M
>Sent: Friday, October 15, 2004 6:27 PM
>To: Joel M. Halpern; [e-mail]
>Cc: Yang, Lily L; [e-mail]; Steve Blake; Alan DeKok;
>Deleganes, Ellen M; [e-mail]; [e-mail]
>Subject: RE: [Forces-protocol] RE: GET/SET in one msg ?
>
>
>Does anyone else have an opinion on this, especially if it is different
>from what Joel suggested ?
>Pls do let us know asap.
>
>I am fine with this, just want to make sure there are no contradicting
>opinions in the Model team.
>
>
>Thanks
>Hormuzd
>P.s. BTW, I don't think I still got a real example on how this would be
>useful.
>
>
>-----Original Message-----
>From: Joel M. Halpern [mailto:[e-mail]]
>Sent: Friday, October 15, 2004 6:20 PM
>To: [e-mail]
>Cc: Khosravi, Hormuzd M; Yang, Lily L; [e-mail]; Steve
>Blake; Alan DeKok; Deleganes, Ellen M; [e-mail];
>[e-mail]
>Subject: Re: [Forces-protocol] RE: GET/SET in one msg ?
>
>That would be my suggestion.
>
>To be specific, I would not have a Query and a Modify message, but
>rather
>would have an Operation message which can carry whatever operations we
>decide we need.  We may make specific rules that say "for sanity, never
>combine X and Y".
>
>I believe this will keep the protocol simpler.
>
>Yours,
>Joel
>
>At 08:40 PM 10/15/2004 -0400, Jamal Hadi Salim wrote:
> >Joel,
> >We are sort of in a rush here to beat a deadline ;-> Give it to us in
> >boolean logic please ;->
> >
> >Did i read correctly that since we may have multiple operations (we
>have
> >been discussing event un/subscribe as something that would appear as an
> >operation for example) then the way to go forward is have GET as an
> >operation?
> >
> >cheers,
> >jamal
> >
> >On Fri, 2004-10-15 at 20:11, Joel M. Halpern wrote:
> > > If we are sure that the only two operations we will ever need are
>GET and
> > > SET, then we could probably simply declare that a message was either
>a GET
> > > message or a SET message.
> > > However, we have had suggestions of INSERT operations, and I would
>hate to
> > > design the protocol so that we could not add other operations later.
>And
> > > some combinations of operations may make sense together (insert item
> > > A.  Add reference to A in item B.  Delete obsoleted item C.)
> > > Thus, I tend to think that it makes sense to structure the protocol
>so
> > that
> > > a single emssage can carry multiple operations.
> > > At the same time, as I said earlier, I would either prohibit or warn
> > > against combining update and read operations in the same
> > request.  Requests
> > > to read, for example to confirm the results of an update, ought to
>be sent
> > > separately so that the FE does not need to worry about the order of
> > > application.
> > >
> > > Yours,
> > > Joel
> > >
> > > At 03:55 PM 10/15/2004 -0700, Khosravi, Hormuzd M wrote:
> > > >No, I don't that's why I asked...since this was coming from Joel's
> > > >proposal.
> > > >I didn't get a good reason from his email either, but it seems like
>he
> > > >would like to have it supported by the protocol anyway.
> > > >
> > > >Joel, do you have any examples for us ?
> > > >
> > > >
> > > >Thanks
> > > >Hormuzd
> > > >
> > > >-----Original Message-----
> > > >From: Yang, Lily L
> > > >Sent: Friday, October 15, 2004 3:52 PM
> > > >To: Khosravi, Hormuzd M; 'Joel M. Halpern'; '[e-mail]';
> > > >'Steven Blake'; 'Alan DeKok'; Deleganes, Ellen M;
>'[e-mail]'
> > > >Cc: '[e-mail]'
> > > >Subject: RE: GET/SET in one msg ?
> > > >
> > > >I don't understand why you would want to do such a thing. Do you
>have
> > > >any example in mind?
> > > >
> > > > > -----Original Message-----
> > > > > From: Khosravi, Hormuzd M
> > > > > Sent: Friday, October 15, 2004 2:04 PM
> > > > > To: Joel M. Halpern; [e-mail]; Steven Blake;
> > > > > Yang, Lily L; Alan DeKok; Deleganes, Ellen M;
>[e-mail]
> > > > > Cc: [e-mail]
> > > > > Subject: GET/SET in one msg ?
> > > > >
> > > > > Hi Folks,
> > > > >
> > > > > We (protocol team) are finalizing some of the msgs and one of
> > > > > the issues which is being discussed is whether GET/SET
> > > > > operation need to be combined in a single msg...(currently we
> > > > > have them as separate msgs). I have never seen this being
> > > > > done in practice i.e. command bundling of GET/SET, but if you
> > > > > guys have some experience/opinions on this, pls do let us know.
> > > > >
> > > > >
> > > > > Thanks a lot,
> > > > > Hormuzd
> > > > >
> > > > >
> > >
> > >
> > > _______________________________________________
> > > Forces-protocol mailing list
> > > [e-mail]
> > > https://www1.ietf.org/mailman/listinfo/forces-protocol


_______________________________________________
Forces-protocol mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/forces-protocol