[Sugar-devel] [DESIGN] Extending activity microformat spec to include optional olpc_activity_size

Daniel Drake dsd at laptop.org
Sun Nov 28 18:03:23 EST 2010


On 20 November 2010 05:59, Anish Mangal <anishmangal2002 at gmail.com> wrote:
> I'm in process of developing an updater for sugar that reads urls (for
> example [1]) which contain information in a format compliant with the
> 'activity microformat specification' [2]. While writing the updater
> code, I realized that I have to make lots of time and bandwidth
> consuming http requests to ascertain activity bundle size; information
> that can easily be included in the microformat embedded in the update
> url.
>
> Thus, I propose to include an optional 'olpc-activity-size' tag in the
> microformat specification, so as to rid the world of extra http
> requests. If anyone has any concerns or suggestions, please speak up
> :-)

I don't buy it.

The updater only has to make such a size query if the activity in
question is an update to one that is installed (or if the activity is
not present on the local system).

Activities are generally megabytes in size, or at least hundreds of
kilobytes. This transfer would follow the size query.

Your proposal would just shave off a handful of bytes.



Additionally:

In my experience, even in OLPC deployments in wealthier nations, the
latency on the internet connections (especially when split between
hundreds of students in a school) is so high that its totally
impractical to consider activities hosted on a remote server. Size
querying or not, its just so painfully slow and unreliable.

The only viable option is to host the activities (and activity group)
locally on the school server. And OLPC provides good infrastructure to
do this, which is in use in various countries.

And when you're on the local school server, size querying delays
become even more insignificant.


If you're wanting to make efficiency improvements in activity
updating, I would suggest tackling the bigger fish such as
http://bugs.sugarlabs.org/ticket/1499

Daniel


More information about the Sugar-devel mailing list