This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author dmgrime
Recipients
Date 2003-08-04.14:34:53
SpamBayes Score
Marked as misclassified
Message-id
In-reply-to
Content
Logged In: YES 
user_id=699265

Thanks Skip.  I see that the ripple effect here is enormous 
(not surprising, given that socket is used by many standard 
modules).  As for #3 - I understand you can use 
marshal.dumps() and then sock.sendall() to achieve the write 
side of a socket-based marshalling transaction, but for a long-
lived connection, marshal.load() on the .makefile()'d file of the 
read-side provided a very nice message-boundary mechanism, 
i.e. - marshal was the "protocol" on the wire...

Obviously (I've already done this in my application) I can just 
use a thin length header on the wire, but it was just very nice 
that marshal accomplished this by itself.  I can also (since my 
platform scope is known, controlled, and limited) create a 
socket.py local to my application which is simply (I don't use 
getfqdn()):

from _socket import *

And leave everything else as-is.  Again, I appreciate the 
attention and effort - I think the ramifications of eliminating 
the "native C" implementation which previously 
backed .makefile() (where supported) are more than just the 
impact on marshal - I see potential performance loss for 
things which rely on it.  In any event, generic object support 
for marshal.load() and .dump() would be great!

Thanks again, Dave.
History
Date User Action Args
2007-08-23 14:15:31adminlinkissue780354 messages
2007-08-23 14:15:31admincreate