At 21:44 14-10-2001 +0800, Federico Sevilla III wrote:
On Sun, 14 Oct 2001 at 08:20, Alan Eldridge wrote:
> Steve, lemme try again.
I'm not Steve (Lord, that is, there are other Steves of filesystem fame,
like Steve Best, for instance), but I'll give it a shot anyway. :)
Not forget there are other Steves posting question _and_ answers to the
list to confuse us.
> For production (business/academic/???) users, people for whom a change
> in filesystem is a flag day that costs $$, a choice to switch to a
> kernel that is only provided by SGI is a tough one.
I tend to agree with you, although the fact that XFS on Linux is GPL'd
should mean at least something (ie: should SGI ever decide to drop support
for it, people can continue as of the latest source code released under
GPL).
That is the main point in which I could switch management. You can always
call up a support company and pay them to keep it going say for a few
thousand euro per month (it's not always full time work).
If five company's do this you have probably all the bases covered.
Perhaps the CVS server provided by SGI can be considered something like,
say, a "donation" (in some ways it may be, in some ways not)?
The paid for the server, they pay for the bandwidth (as Steve L once
mentioned it was around 300GB in a few days).
> I don't know how I'd justify such a thing. The old question of
> support/maintainance/longevity is a real tough one to wave away in
> that case.
I don't know if Seth Mos ever was with SGI, but I don't think he's with
SGI anymore. I'm not really sure. What I'm positive about, though, is that
Seth is still very active in the list and handles the upkeep of the FAQ,
which is hosted in SGI. There were also those SGI layoffs that even got
Slashdotted, specifically because a number of XFS developers had to go. I
do believe they're somehow still active. Not quite as much as when they
were paid to be given that they have other day jobs to attend to, but
still...
I have never worked with SGI, it does not seem likely that I ever will work
for SGI (offers per private mail ;-), and I do this entire voluntary.
I work for Coltex Retail Group BV in the netherlands as a system
administrator and we use XFS at work in production systems and thus I need
support. Half of that support is giving back to the community. The other
half is supporting the systems myself. I feel confident enough that if
something ever comes up we can just call a company like StoneIT in the
netherlands and ask them for support if we need it. For a price ofcourse,
but then again support never comes free.
> What would you say to such a user in making the case to use an XFS
> kernel, when technical arguments alone won't do?
It's GPL'd. Worst case scenario we can all pitch in and try to understand
the code from wherever it left off. And it's not like the XFS code is in
its infancy. It's quite mature (very well tested on IRIX, and the Linux
port is awesome), so we won't have to reinvent things anyway, just
understand and maintain.
Remember, there are a lot of companies like Linuxcare that do just that.
Offer support on stuff that otherwise would be lost. Since XFS has great
value I don't see this happening.
This doesn't say that XFS won't or shouldn't be pushed to the mainstream
kernel. But it's a hell of a lot of work, and I don't think the SGI guys
are being paid to focus their lives on cutting out all the duplication.
They _can_, but we all only have so much time. If there's something the
community can do it's probably pitch in. And I know for a fact that the
SGI guys listen. We've already seen help from that guy from NEC who was
listened to when he some stuff for xfsdump, I think it was (one of the
userland tools).
It just takes time and a bit of help from everyone. It is not really that
difficult. I answer questions on this mailing list in my free time and some
of them at work in between moments of peace and quiet. However due to a new
server at work and the stampeding/impeding Euro conversion tricks I have
less time but I already see other people stepping in and answering
questions. Things that make you feel warm and fuzzy inside :-)
And I do believe getting XFS to 2.5 and then 2.6 is on the top priority
todo list of Steve Lord and company. :)
We are not forgotten, Alan is actually testing and feeling about XFS every
know and then. He just does not tell us. We are being noticed, don't worry.
Cheerio, off to the coffee
--
Seth
Every program has two purposes one for which
it was written and another for which it wasn't
I use the last kind.
|