On Thu, Mar 22, 2001 at 04:28:18PM +0100, Wolfgang Solfrank wrote:
> Hi,
> 
> so I decided to finally look why I got checksum errors on some, but not
> all files, when trying to build some packages from source.
> 
> Turns out that the problem is that I'm using a proxy to fetch the data,
> and that it adds CR to the end of the lines on text files.
> 
> Looking deeper into ftp source, it seems there is an inconsistency here.
> While ftp, when not using a proxy, goes out of its way to fetch a
> file in image mode by default, it doesn't do so when going via a proxy.
> 
> Furthermore, without a proxy, it even tries to deduce the user's intention
> by scanning a trailing ";type=i", including stripping this extension from
> the resulting filename for saving onto the local disk.
> 
> In contrast to that, when using a proxy, and trying to get the unmodified
> file you add a trailing ";type=i" to the url, ftp doesn't strip off this
> extension.
> 
> Shouldn't ftp behave the same independent of the use of a proxy?
> 
> Currently, scripts trying to make use of ftp (like the package framework),
> would need to look whether the ftp_proxy environment variable is set, and
> recover from the different behaviour (which the package framework currently
> doesn't do).

I'd prefer it, from a pkgsrc PoV, if it was totally transparent
whether or not a proxy was in use.  OK, sloping shoulders time, but
why ripple inconsistencies up the foodchain, when it should be done in
the lower layers?

i.e. I agree with the rhetorical question:

> Shouldn't ftp behave the same independent of the use of a proxy?

Regards,
Alistair