patch: arla 0.5 on Linux/glibc2

[Assar Westerlund]

[ I'll only answer the stuff that Love hasn't already or the parts
where I don't agree with him ]

Brad Keryan <keryan at andrew.cmu.edu> writes:
> It also changes the linking of libroken from -lroken to
> .../libroken.a because otherwise gcc finds the libroken.a from kth
> krb4 and uses that; is there a better way to make it link against
> the right library?

The correct way of doing this is to have -L for the local directories
first and then -L/usr/athena/lib.  This applies not only to libroken
but other libraries as well.

> I still haven't gotten the kernel part working with Linux 2.0.33/x86 and
> have not had the time to take a closer look.

It's working (kind of) on 2.0.33 and work is on the way for 2.1.

> Also, what does everybody think of the idea of an ftp-type interface
> to arla?

I find myself using `arlad -t' quite seldom these days and only for
some specific debugging.  The `real' filesystem is much better.

But if you find it useful, go ahead.

> I think it would be useful to extend the "arlad -t" interface to
> allow "put" and "get" commands and other niceties. Would it be better to
> make this a separate program from arlad, with a per-user cache? 

I think it's much simpler to have it in the same program.  Just add an
option for per-user cache to arlad (changing the chdir).

> Also, does any of arlad need root privileges for anything other than
> directory permissions on /usr/arla/*?

No, it only requires permission to write in the cache directory.

/assar