[isf-wifidog] Gateway development
acv at miniguru.ca
Lun 7 Nov 17:28:03 EST 2011
On Mon, Nov 07, 2011 at 04:26:23PM -0500, Genevive Bastien wrote:
> Thanks for this work. Unfortunately I don't personally have the time or
> the busy network to test it out, but if need be, we may take a few nodes
> at île sans fil to test this stuff.
> As for protocol V2... it would have to be designed for today's needs.
> The specs in the wiki are already quite old.
I don't think the specs are much relevant.
> One important thing is really to avoid the 1 request/connection/minute +
> 1 request/node/minute thing which puts a lot of load on the server. 1
> request/node/minute should be enough to do all there is to do.
This is good. And it makes sense to implement, this issue is the
current code is a loop that refresh the client list and look at all of them
as it refresh them. We can split that into two process, one that refresh the
usage timer and one that check for session validity.
> Also, if this new protocol could find a way to bypass the Apple
> auto-login thingy, it would be great!
I think that's separate from the wire protocol.
> Also, no extra libraries should be needed as it should keep working with
> extra low power devices.
We'll probably need a parsing library but it should be embedded
friendly (i.e.: expat or one of the super light JSON libs) and not huge
> That said, I don't know enough of the gateway to work on its
> implementation, but I'd be glad to make AuthPuppy work with this new
> protocol when it is available!
I would thing the auth server and user needs are what will drive
the protocol requirements and not the other around (the current scheme is
what you get when the gateway drives the protocol!)
I'd like use cases and requirements to work from, if anything.
> On 11-11-06 01:56 PM, acv wrote:
> >I've picked up a bit the gateway code and tied up some loose ends (as
> >in off-by-one errors in boundary checks, unchecked strcpy(), etc.) over
> >the week-end. I have a bit of time now to improve things on that front.
> >For now I have put my changes at https://bitbucket.org/acv/wifidog-alex
> >(I know I have commit access to the SVN, but I much prefer the branching
> >model of DVCS, especially for potentially destabilizing changes.)
> >I would be highly appreciative if some hardy souls out there could pick
> >it up and test it (in controlled form) in networks where there's a
> >significant load. My test setup is a Ubuntu box (vm actually) and does
> >not include many devices so I can't really stress test.
> >I'd like to maybe look at implementing the fabled v2 protocol on there
> >(that is, the one that doesn't exists) if some of the AuthPuppy folks
> >want to hash out something.
> >Alexandre Carmel-Veilleux
> >WiFiDog mailing list
> >WiFiDog at listes.ilesansfil.org
> WiFiDog mailing list
> WiFiDog at listes.ilesansfil.org
-------------- section suivante --------------
Une pièce jointe autre que texte a été nettoyée...
Nom: non disponible
Taille: 195 octets
Desc: non disponible
Plus d'informations sur la liste de diffusion WiFiDog