>
>Thanks for the report. I''m probably going to rework the entire
channel
>implementation during the rewrite/refactor that I''m in the middle
of,
>which should help improve SFTP performance.
Yes I''ve been reading the effort from your blog. Seems to go quite
well.
(Note, though, that Net::SSH>will never be as fast as any of the OpenSSH programs...in a Ruby vs. C
>context, C will almost always win.)
>
This is understandable in general. But in this case the speed loss,
is
quite big. Maybe chunk_size or something could be reason.
>I don''t think the acculating_data_packet is a problem (at least, in
this
>case). The warnings are because I don''t set it to an initial value
(in
>the constructor, for instance).
Yes, there is plenty of uninitialized variable warnings when running
net-ssh in debugging mode. How ever the @accumu... variable is
never set. Or not in my debugging.
>
>One caveat: very little effort has been put into improving performance,
>and this trend will probably continue until Net::SSH is
>feature-complete. Once all the features are implemented, I''ll go
back
>through and look for ways to optimize. Basically, this means that the
>current Net::SSH implementation is more of a proof-of-concept than a
>real production-worthy SSH implementation.
This is fair enough.
>
>- Jamis
>
>--
>Jamis Buck
>jgb3@email.byu.edu
>http://www.jamisbuck.org/jamis
>_______________________________________________
>Net-ssh-devel mailing list
>Net-ssh-devel@rubyforge.org
>http://rubyforge.org/mailman/listinfo/net-ssh-devel