[Telepathy] Spec changes: tube API

[Guillaume Desmottes]

Le mercredi 25 juillet 2007 à 16:40 +0100, Dafydd Harries a écrit :
> Ar 25/07/2007 am 09:55, ysgrifennodd Simon McVittie:
> > On Wed, 25 Jul 2007 at 10:24:56 +0200, Guillaume Desmottes wrote:
> > > All these comments are fixed in
> > > http://projects.collabora.co.uk/~monkey/spec-stream-tubes-smcv/ 
> > > 
> > > Rob, did you finish your review of this ?
> > > Any other comment/suggestion ?
> > 
> > In a text message Rob sent me at Guadec, he wrote:
> > > Feel free to merge stream tubes, but split ip4 and 6 I think.
> > 
> > If I remember correctly, Rob's suggestion was to have the address types be
> > as follows:
> > 
> > Tube_Address_Type_Unix: 'ay' containing filename
> > 
> > Tube_Address_Type_Abstract_Unix: 'ay' containing the sun_path member of
> > struct sockaddr_un, which is an array of char of length <= UNIX_PATH_MAX
> > (FIXME: do we include the leading NUL or not? I'd be inclined to include it)
> 
> I think the leading null is an artefact of the sockets interface rather than
> something inherent and therefore should not be copied to our interface where
> we don't have any compatibility/misc concerns.

> > Tube_Address_Type_IPv4: a struct '(ayq)' where the ay has length 4
> > 
> > Tube_Address_Type_IPv6: a struct '(ayq)' where the ay has length 16
> 
> This is moving a burden from the CM to the client. I'm not saying that's good
> or bad, just saying that we should judge this depending on whether we think
> that moving that burden is justified or not.
> 

Any progress on these 2 points?

I don't have any opinion about them. I'd just like to have this spec
merged this week so I'll be able to update Salut and Gabble before my
holidays.


	G.