[Novalug] thunderbird and ldap

Mark Smith mark@winksmith.com
Tue Jul 8 01:13:12 EDT 2014


that statement makes perfect sense, but it seems that if a program will
crash when it gets an answer in 200ms vs. 2ms there are some serious
problems.  my gut feel is that the time difference between two small time
spans can't make the difference between running and crashing.  ether way,
dumping core when a response takes seconds vs ms is still broken.
my guess is there's some broken logic in there.

before you ask, i care enough to send this email, but not enough to
actually look at the code, so who knows.  :-)

On Mon, Jul 07, 2014 at 09:02:16PM -0400, Derek LaHousse wrote:
> nscd caches the answers.  So when it asks, it gets an immediate
> response, as opposed to having to network to the ldap server.
> 
> On Tue, 2014-07-08 at 10:00 +1200, Mark Smith wrote:
> > On Mon, Jul 07, 2014 at 12:16:02PM -0400, Peter Larsen wrote:
> > > Can anyone come up with a use-case where access to the local users would
> > > be needed?
> > 
> > upon reflection, it's reasonable to open the file.  my nsswitch.conf
> > file clearly said "files ldap".  if i were to intrepet that, i would look
> > for the user in /etc/passwd first then in ldap.  i can clearly see that
> > it opened the password file from system call trace, but i can't see the
> > logic performed on it from that point.  it could have been looknig for
> > the user in the password file and then failing that, looked in ldap.
> > this is exactly what it's supposed to be doing.
> > 
> > i wonder how nscd makes this better though.  that suggests something to
> > do with timing, but that's absurd.  ldap response is "too long" but nscd
> > is fast enough?  i would imagine that would be a tiny enough delay like
> > in the order of ms not enough to make a difference.
> > 
> > i think it's just a bug.  

-- 
Hei konā mai
Mark Smith
mark@winksmith.com



More information about the Novalug mailing list