launchpad-dev team mailing list archive

[Robert Collins <robert.collins@xxxxxxxxxxxxx>]

On Tue, Aug 24, 2010 at 8:31 AM, Sidnei da Silva
<sidnei.da.silva@xxxxxxxxxxxxx> wrote:
> On Mon, Aug 23, 2010 at 4:45 PM, Robert Collins
> <robert.collins@xxxxxxxxxxxxx> wrote:
>> This has two characteristics that what I described does not, and both
>> are constraints.
>> Firstly, it adds another client round trip. Thats (optimistically)
>> 300ms from here, and also, importantly, upwards of 200ms in Asia,
>> where we have a lot of commercial users.
>
> Presumably you'd be waiting for more than 300ms to get an actual
> response, otherwise you would just do it within the same transaction?

See back up in the thread. This is a variable. Sometimes it takes
10ms, sometimes it takes 5000ms. Arguably thats a bug in *this
particular case*. In general though we expect to have things which are
often fast and sometimes slow.

>> Secondly, the Landscape implementation uses a new host name, which
>> adds an SSL handshake if no TCP connection is open (common for only
>> occasionally used services), and even key exchange if the key is out
>> of the cache.
>
> Not sure where you got this one from. It uses the same host name.
> Requests to '/ajax' are simply forwarded to a different server in the
> backend by way of mod_rewrite.

Ah, incorrect information - thanks.

>> Thirdly, you need to render long poll responses in a different
>> appserver environment than your regular one, which adds programming
>> and programmer friction.
>
> It's slightly different, which adds some friction. Justifiable though, IMO.

What is the justification?

-Rob