[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnu-arch-users] Re: tla on nfs only

From: Aaron Bentley
Subject: Re: [Gnu-arch-users] Re: tla on nfs only
Date: Sun, 14 Mar 2004 13:48:52 -0500
User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.6b) Gecko/20031205 Thunderbird/0.4

Stefan Monnier wrote:

Finally, this is a Free Software project, and part of that freedom is the
ability to control how your tools work.  The change in question is not
difficult to make.  Lobbying volunteers to make the changes you want doesn't
seem a very nice way to operate.

Well, it's not quite the same:
- I can't proselytise the same way if I have to tell people "oh don't use
 the official version but use mine instead".
My suggestion is "write it first, then get it included". It's quite conceivable that checking revlibs for corruption could become optional, but it's not going to until someone who cares does the work. And that someone is not me. It seems to be you. Now that you know where to disable the checks, you need to implement a way of making it configurable. I'd suggest a =nocheck file in the revlib directory, and an extra option "--no-checks" for "library-config". That way, if you have some libraries on NFS and others locally, you can disable checking for the NFS revlibs only. Probably be a good idea to mount that NFS partition read-only, too.

- When writing support code (like an Emacs front-end), I can't assume that
 people will use my hacked version.

NFS is a reality at many places (Samba at others, tho I'm not sure it's
much better in this respect), tla needs to deal with it IMO.
NFS is a reality at my workplace too, it's just that we don't use it for storing home directories. tla falls down only when users are denied reasonable access to local storage. It does a great job with *archives* in NFS.

I think the core of the problem is that too many files of the revlib are
checked every time.  If tla were more careful to only check the files it
actually looks at, then it would have much more freedom to use other kinds
of checks that don't depend on inode or device numbers.
See my last email. In order to know which files changed, we must check all of them.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]