[OpenAFS] Re: GiveUpAllCallBacks callers
Wed, 15 Dec 2010 10:58:48 -0500
Andrew Deason wrote:
> On Tue, 14 Dec 2010 10:47:22 -0500
> Derrick Brashear <email@example.com> wrote:
>>> b) Do something with the version string in the meantime
>> i'm not wild about this.
> I kinda would want to explore this more. This doesn't seem worse than
> the GetStats approach, and makes a certain amount of additional sense.
> It doesn't catch all cases, but it catches the common case where someone
> builds the code without changing anything or doing anything special.
> It also does not affect other potential implementations and avoids
> implicit relationships between otherwise unrelated parts of the
> protocol. At least, in theory; it obviously ties the RX version response
> to GUACB behavior but I think that makes it very very obvious that it's
> a minor implementation-specific quirk.
> I'm unsure of the reliability of rx version packets, though. And the
> existing libraries I think make this more annoying and involved than a
> regular RXAFS call, which makes this much less appealing...
This does catch the most common case of a fileserver running a vanilla
openafs release. I have seen people append patch level info to the
version string, but I don't think it is very common to change the
major.minor version numbers. This approach maybe the best of a bad
set of choices of we do not want to release a client that uses a
known buggy rpc for fileservers which were and may still be in production.
> An RPC with a version vector or something more reliable than a text
> string has already been discussed and is presumably going to be in the offing.
So, if I understand, then the version string check (or get-stats rpc check) is
meant to be a stop-gap?