launchpad-dev team mailing list archive

[Deryck Hodge <deryck.hodge@xxxxxxxxxxxxx>]

On Thu, Nov 10, 2011 at 8:37 PM, Huw Wilkins <huw.wilkins@xxxxxxxxxxxxx> wrote:
> On 11/11/11 13:33, Robert Collins wrote:
>>
>> On Fri, Nov 11, 2011 at 3:26 PM, Huw Wilkins<huw.wilkins@xxxxxxxxxxxxx>
>>  wrote:
>>>>
>>>> Implementation wise we don't have node.js on the servers at the
>>>> moment. I suggest testing it and if its fast enough in 'client side
>>>> mode' just using that for now (perhaps file a high priority bug for us
>>>> to permit server side handling in the future).
>>>
>>> Thanks Rob. I'll look into that.
>>>
>>> I have seen at least one python implementation of the compiler in the
>>> past
>>> (I remember it wasn't feature complete at the time). I assume if that
>>> fits
>>> our needs we could use that in place of the node.js compiler.
>>
>> I have no objections. OTOH if its pleasantly fast on the client, it
>> would be simpler to have just one compiler (the .js one) and run it
>> only on the client. Simple is nice in its own right ;)
>>
>>
> I also am happy either way, I'll plan to implement the client js compiler
> and I'll investigate what performance gains we get from server-side
> compiling and if it's trivial we can stick with the client compiler.
>
> It's also possible our CSS builder won't be able to handle less files (in
> which case we'll have to do server-side compiling first), but I guess I'll
> find out.
>

Doing this in the browser makes me nervous.  I can't imagine that we
can do this without seeing some page performance hit.  But I welcome
real testing and data to prove me wrong.  Can we please use real
profiling/timing data from a web dev console against real pages and
not just a finger in the air, I don't see any problem with it, kind of
test?

I don't mean to be harsh here, but we already struggle with js slowing
pages down in places.  This client-side compile seems completely
unnecessary to me, though I do understand concerns Robert has about
adding a dependency for node.  I'm happy for you guys to investigate
client-side use, but I can't help but be skeptical of it.  And I'd
just like real proof that it's not any slower.

Cheers,
deryck

-- 
Deryck Hodge
https://launchpad.net/~deryck
http://www.devurandom.org/