[Users] [Bug 2828] Use MD5 digest for socket name

[Colin Leroy]

On 01 December 2012 at 20h26, Rich Pieri wrote:

Hi, 

> > So, CVS now has rc_dir canonicalized. Two issues with that patch
> > though: 1) unified diff please  
> 
> I'll work on it this coming week.

Is that superceded by your next mail (ID
20121202000300.00006610 at unknown)? So that I know whether to do it or
not.

> I've been busy the past couple of days. Perhaps you heard about the
> city-wide power failure in Cambridge, MA on Thursday afternoon? My
> mail server ate it's system drive because of that despite the server
> being on a UPS and being shut down cleanly (it seemed). I spent
> Friday afternoon and evening building a new mail server and restoring
> my users' mail from backups and the salvaged data drives.

Yeah, servers tend to suicide themselves for no apparent reason at the
worst times possible... That's annoying as hell! 
I tend to put SSDs in the mission-critical servers now, they seem to
handle age better than disks, at the expense of being not so big.
 
> > So, TL;DR for 2) is that the normal socket for normal RC dir must
> > not change name until the sun goes red giant on us :)  
> 
> Does it make sense to test the existence of both sockets and do
> the right thing if either exists?

Yes :)

> > The claws-mail-uid/md5hashsocket which has been proposed on the
> > users list solves this problem, as clashes with existing
> > claws-mail-uid socket files.  
> 
> I proposed this. I was told, "[i]f it's to be per-user, it should just
> be XDG_RUNTIME_DIR."

Well, that was a good idea (your proposition). The problem with
XDG_RUNTIME_DIR is that it won't work for sure and we have to have a
failback in any case. That makes a lot of logic for not very much
benefits.
-- 
Colin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 190 bytes
Desc: not available
URL: <http://lists.claws-mail.org/pipermail/users/attachments/20121202/a2325164/attachment.sig>