[issue3566] httplib persistent connections violate MUST in RFC2616 sec 8.1.4.

[Demian Brecht]

Demian Brecht added the comment:

My apologies for the delay, but I've now reviewed the proposed patch. With a fresh outlook after taking a bit of time off, I'm not sure anymore that this is the best way of going about solving this problem. The main reason being that we now have two errors that effectively mean the same thing: The remote socket has encountered some error condition.

I understand that ConnectionClosed was added to maintain backwards compatibility with the BadStatusLine error, but I'm beginning to think that what really should be done is that backwards compatibility /should/ be broken as (in my mind) it's one of those cases where the backwards compatible solution may introduce just as many issues as it solves.

The root issue here (or at least what it has turned into) is that BadStatusLine is incorrectly raised when EOF is encountered when reading the status line. In light of that, I think that simply raising a ConnectionError in _read_status where line is None is the right way to fix this issue. Not only is it consistent with other cases where the remote socket is closed (i.e. when reading the response body), but it's removing the need for the addition of a potentially confusing exception.

I'm not 100% what the policy is around backwards introducing backwards incompatible changes is, but I really think that this is one of those few cases where it really should be broken.

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue3566>
_______________________________________