Hi

Sometimes when I'm bored, I look around the internet for various curl-related
stuff that happens where I normally don't look.

One area where I've found curl usually gets a lot of unjustified badmouthing,
is for example when used by 'urpmi' (the Mandrake installer program) and
similar programs.

The reason for this is that the program uses curl or wget for downloading
packages off the net. It doesn't do anything different in the program, it just
invokes the specific download program to get a remote file.

While this may look like a good idea to many people, it doesn't get the
intended behaviour. The most typical case is when people attempt to download a
file off a very busy FTP server: wget has an internal retry functionality that
makes it retry ftp transfers if the server says it is temporarily busy, while
curl has none and exits immediately on all kinds of errors (assuming that the
one that use curl takes care of retrying or not).

This makes curl look bad compared to wget in certain cases and I don't expect
all users of these programs to fully grasp why this is so (and many probably
don't care), and apparently the authors of these programs aren't caring enough
to do much about it.

I'm considering adding a new --ftp-retry option that would do exactly this: if
the FTP server returns a trancient error when we attempt to login, we sleep a
while and retry again later. We double the sleep time for each attempt and we
do only a limited number of attempts.

Is this a good idea? I honestly can't think of many drawbacks, other than it
pollutes the code somewhat with some rather FTP-specific stuff and an extra
loop.

If it is a good idea, would we also need a --ftp-retry-count (to set maximum
number of retries) and --ftp-retry-time (to set maximum time allowed for
retries)?

What do you think?

-- 
      Daniel Stenberg -- http://curl.haxx.se -- http://daniel.haxx.se
       Dedicated custom curl help for hire: http://haxx.se/curl.html