On Sat, 29 Oct 2005, Mehmet Fidanboylu wrote:

>>> 3. Has anyone attempted to replace CURL's timer mechanism? How hard is
>>> that?
>
>> I don't understand what you mean with the "timer mechanism". Exactly what
>> are you considering?
>
> CURL has internal timers to determine metrics such as connect time,
> pre-transfer time etc..

I wouldn't really describe them as "internal timers". libcurl just remembers
at what time everything starts and then it stores the time (again) at various
checkpoints.

> I was just curious to see if anyone attempted to replace these timers to get
> more sensitive data..

That would mean you need to call other functions to get the time.

> I am particularly concerned about Windows NT performance of these standard
> timers..

And what timers are they that you are concerned with? You don't seem to even
know what functions libcurl use.

> I have read somewhere that their precision is only around .1 sec..

The precision of what?

The windows version of libcurl uses timeGetTime() by default, which according
to some sources I found offers a 1ms resolution on some Windows versions but
on NT/2000/XP it returns 10 ms resolution.

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html