On Jun 23, 2009, at 3:34 PM, [email protected] wrote:
Folks,
For a while, our transport system has not really been adequate to
address all of the goals for the pkg project. The python libraries
that
we have been using are capabable for basic functionality, but fall
flat
when we try to perform more complicated or performant network
operations.
The following webrev converts our transport from using the
urllib/httplib in python to PyCurl, which is based upon libcurl.
http://cr.opensolaris.org/~johansen/webrev-xport-1/
Some other comments after poking around with the transport api today:
* I noticed a ProgressTracker isn't supported when using get_url() (or
rather, fetching single resources instead of multiple). Given the
size of some our manifests (2.8M for SUNWjruby), I think tracking the
download progress would be helpful :) In fact, updating through
packagemanager for a new release will trigger a download of at least
65 megabytes of manifest data!
* It would be good if stats.dump() included the total bytes
transferred (which I ended up hacking in my local workspace so I could
compare tools.gzip.on to off)
* Maybe stats.dump should be changed to use misc.bytes_to_str and just
label the KB/sec column 'Speed'
* It would be nice if the additional stats that the object has as
private properties (such as __bytes_xfer) could be exposed
Cheers,
--
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
