>>>>> "D" == Daniel Stenberg <daniel-curl_at_haxx.se> writes:

    D> On Fri, 10 Jun 2005, Gary Lawrence Murphy wrote:

> I'm experiencing serious, strange and sporadic data corruption using
>> PUT requests from curl-devel-7.10.4-1 and
>> curl-devel-7.10.6-6.rhel3

    D> And when you try the same program with libcurl 7.14.0, what
    D> happens then?

    D> You'd have to come up with some pretty strong and convincing
    D> arguments to make me even consider digging around in such old
    D> releases for a bug that very likely is already fixed.

But you have to admit: This is a pretty cool bug!

Think of it: Libcurl 4.10 can deliver data from the /future/ ...
that's a pure-software digital exploitation of multidimensional String
Theory! :) Think of the millions to be made inserting, say, future stock
quotes into random nigerian scam emails! I'd say that's a pretty
compelling argument in itself.

But, ok, seriously ... you have a point :)

These are the releases considered "current" for RH9 and RH FC2
respectively, pretty 'old' by Linux standards, and yes, perhaps the
bug is fixed and yes, I will be building libcurl from current sources
if I can't find a more expedient way to /avoid/ the bug.

But it's a damn fine bug nonetheless. I haven't looked at the curl
code in any depth to really say it cannot happen, but I cannot
conceive of any model where this bug /could/ exist. If it switched
from Doc A to Doc B mid-stream, that I could comprehend, but to insert
future doc headers into a document already in transfer and before the
future doc is even known? That's freakin' awesome ;)

-- 
Gary Lawrence Murphy <garym_at_teledyn.com> ==============================
www.teledyn.com - blog.teledyn.com - irish.teledyn.com - sbp.teledyn.com
====================== The present moment is a powerful goddess (Goethe)