>> I wonder if we could have another approach. We could detect the PASV
>> collision and repeat the PASV commands (it seems that IIS accepts that),
>> it would cause the slow-down only in case of actual collision. I well
>> understand that this is "IIS bug handling" but IIS is really spread and
>> having libcurl just inverting silently the data when you try to download
>> two files in parallel from an IIS server is not cool.
>
> Since you're the first person ever to report this, I don't think this is a
> common problem for libcurl users and I don't think we should bend over
> backwards to support it.
>
> If you can come up with a *nice* approach that isn't excessive for
> conformant servers, then I think we can consider including it.
should I assume that detecting the collision and repeating PASV is -not-
nice?

Armel
Received on 2006-08-24