Clash detection is really hard to do right. I need a lot of
convincing before it can ever be re-enabled by default.
Clash detection has serious conceptual problems.
It is very difficult for emacs to tell if two filenames used by
different clients really clash or not - e.g. two clients editing
/etc/passwd will clash only if they are on the same machine.
In general, clients are going to have separate file name spaces.
clash detection requires a directory writable by every client. In
many environments this cannot be provided or is insecure. In
corporate environments, tools server directories are mounted
readonly.
Even if the global directory is writable, it may be NFS mounted by
thousands of clients worldwide. This creates unreasonable contention
for global, unreliable NFS file locks.
In the case where it does work reliably (single-user machine, no
network filesystems) is also the case where it's the least useful.
A feature that only works intermittently, causes perceptible slowness
on all file operations, and fails to scale well is a feature that
needs to be explicitly enabled by the user.
Keep in mind my vision for the future: someone provides a binary
installed XEmacs that is used potentially by everyone worldwide simply
by setting their PATH. I implemented such a `non-distribution' of
XEmacs in 1993 at IBM. I had users in California, Toronto, and Sweden
using the exact same binary. Do we really want these users to deal
with mutual clash-detection?
Martin
I say NO! to enabling global clash-detection.