[sldev] Certified HTTP Project Update - Oct 25

[Dzonatas]

Hi!

Ryan, I believe you, LL, and your sup have blindly stumbled upon a 
unique protocol, and it maybe just not as clear due to the higher-level 
presentation of it with use of rather robust Internet features.

I came upon this conclusion after I looked over the design of two other 
items that have approached the table. The items are remarkable, and I 
wondered what was needed to complete a system for it.

Here are the two items of attention, that you probably already seen:

[sldev] Proxy Server and UDP Cache Server
http://lists.secondlife.com/pipermail/sldev/2007-October/006227.html

[sldev] AWG: Geometric Content Creation
http://lists.secondlife.com/pipermail/sldev/2007-October/006240.html

Now, I know LL knows UDP pretty well and kinda has a notion to 
completely switch over to more HTTP based systems, but I believe LL had 
it right with UDP design the first time except the packet itself was not 
robust enough.

My conclusion is that the UDP packet needs to be redesigned to make it 
dynamically persist and duplicate as needed. It would be kinda 
TCP-HTTPish, but there are no URLs involved (in mind). I can think of a 
another RFC that is well-know and kinda similar to the idea, but I don't 
want to point it out to cause an unintentional red herring.

I know you wondered how can it it work with the URLs, but this is 
simple... too simple... instead of URL/UUID (or such), there is a 
discrete UUID|UUID format.

I ask to reconsider the UDP stacks that LL already has made, and you'll 
notice, with the two items above being implemented at a much 
lower-level, that there is a lot of bloat middle-ware that causes 
completely useless bottlenecks.

We know that LL needs a solution based on well-known implementations. 
Here you already have "running code" listed here that compliments each 
other nicely to get this done. (very nicely!)

I might have to change my answer for what has been known as practical 
limits for UDP with the questionaire sent me! (Phoenix knows...)

++ Dz

P.S. There was a heads-up slashdotted recently of a company that ganders 
over a similar flow-based internet, but we already know these ideas. =)


Ryan Williams (Which) wrote:
> In the past week, Sardonyx worked on some important details, making
> eventlet.httpc handle HTTP status codes that it didn't handle before,
> and adding unit tests.  I ended up working on other stuff so I appear to
> be the lazy one.  :-)    Development is picking up again today as you
> might see from our commits:
>   http://svn.secondlife.com/trac/certified_http/timeline
>
> Overall view of the project in PJIRA:
>   http://jira.secondlife.com/browse/CHTTP-1
>
> Reference Links:
>   http://wiki.secondlife.com/wiki/Certified_HTTP_Project  (includes a
> quickstart for getting the code and trying it out yourself)
>   http://wiki.secondlife.com/wiki/Certified_HTTP
>   http://wiki.secondlife.com/wiki/Certified_HTTP_API
>
> -RYaN
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html
>
>
>   

-- 
Power to Change the Void