This is the mail archive of the glibc-bugs@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug libc/1936] Improvement to rcmd


------- Additional Comments From sourceware at zacglen dot com dot au  2006-04-23 23:22 -------
(In reply to comment #2)
> The argument about robustness is bogus.

The statement that the argument is bogus is bogus.

The main reason for the change is because the attempts to reuse the
same port number causes high delays.

> We will always iterate over all possible ports if necessary.

Ok, for those people with lots of time to waste.

> The only thing which could be said is that we try
> to many times the ports which might be crowded by other rcmd calls and this
way cause delays.

And that is exactly what has been said.
It is really the only reason for the change.

> 
> But a limited amount of randomization will only help a bit.

And a bit of improvement is all that is needed.

> The next person
> will come and complain that the randomization is too small.

But you will tell them that they shouldn't be using rcmd, won't you?
Besides, you cannot quantify randomness by size.  So you could just
tell them to get lost because random is random no more no less.

> But sooner or later
> we'll run into problems with ports which are explicitly requested by
> applications and who then couldn't run.

Perhaps you haven't heard of registered ports?

> Any change in the way ports are tried
> will increase the chance that this can happen.

Absolute and utter crap.

> 
> Combine this with the fact that nobody really should use rcmd() etc the best
> solution is to not change anything.

Oh really?  I suppose nobody should use Microsoft Windows either.
But surely one persons opinion as to what is a "correct" thing
to use or not use should not be used when discussing bugs.
There is nothing wrong with rcmd.  And there are no rules wrt
what one should or should not use.

>  Maybe this will entice people to refrain
> from using rcmd() which is the best outcome one can hope for.

Oh yeah, the best excuse for a bug I have heard yet ... we are trying
to stop people from using the software because they shouldn't use it ...
Imagine Toyota explaining that it is ok for the wheels to fall off your
Echo because you should be buying a Lexus anyhow.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=1936

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]