Re: [Arx-users] Blog entry: Selecting an SCM

[Peter Busser]

Hi!

> Ah. I know the git folks had some fancy stuff going before they threw it 
> all out and did packing. I don't think they've put fast http pulling 
> back in yet. My memory was that they were talking about parallel pulls, 
> and some http 1.1 (?) feature where you could request multiple pages 
> with one get command? I wan't paying close attention so I could be all 
> wrong.

I think I know which fancy feature you mean. A traditional HTTP request is like
this:

- open socket
- send request1
- read reply1
- close socket
- open socket
- send request2
- read reply2
- close socket

And that for every object, no matter how small it is. The ``fancy'' HTTP 1.1
feature you are probably refering to is AFAIK called persistent connection or
is it cached connections?. Anyways, it works like this:

- open socket
- send request1
- read reply1
- send request2
- read request2
... more requests
- wait some time to see if more requests come
- close socket

This saves the time to setup and close a socket. On slow connections, the time
to do that can be noticable, especially with lots of small objects.

Groetjes,
Peter.