Re: CAB-for-MiNTNet overlay 1.83

[chu@platinum.com (Howard Chu)]

  >At the moment it's undergoing a major overhaul. I have had to completely
  >redesign my URL caching scheme to accomodate authorization/authentication
  >and cookie info. At the same time I'm adding the other major missing piece
  >of HTTP/1.1 - reusable connections. This feature will allow multiple request
 s
  >to go over a single TCP connection, which will boost throughput tremendously
  >for pages with a lot of images loading from the same server. My tests so far
  >show that the typical HTTP 1.1 server times out the connection after 10
  >seconds, so this won't speed up access to a particular server if e.g. you
  >get a page, spend a while viewing it, and then click on another one of its
  >links. At the same time, I'm restructuring the code so that the HTTP-specifi
 c
  >management is separate from the MiNTnet-specific code. This should make it
  >easier for STiK/STiNG hackers to adapt my HTTP/1.1 support. (This overhaul
  >was long overdue; up till now I was just poking and prodding at Dancer's
  >original programming structure. It probably suited his mentality just fine,
  >but it's time to really make it the way I think it should be...)

  Is there any planned release date for the new ovl? Mainly I'm very interested
  in authorization!

Release dates? You must be joking... This is not a commercial project, it's
something I do in my spare time. Of course, I'm interested in finishing it
as quickly as possible, but there are other things competing for my time...

At this point I've pretty much finished the restructuring, and I've got
reusable connections working. I've changed the request version to HTTP/1.1;
I have just finished re-reading the entire HTTP 1.1 protocol spec and I've
got pretty much every required feature working now. At least, as much as CAB
will allow. The 1.1 spec goes into great detail about how a cache should
behave, but all of that is handled by CAB itself... The 1.1 protocol also
allows retrieving sections of a URL, by specifying byte-ranges. I don't
see any way for CAB to do this, so that's not implemented here either...
There are a lot of tiny little details like that which the current CAB design
ignores...

Just fyi, I've spent almost 4 days straight on this, since it was Thanksgiving
holiday for us. It's been a long weekend of hacking. I had intended to get it
all finished this weekend, but it looks like not. I still have to finish the
cookie code and the digest-based authentication.

Rewriting the URL parser took a lot longer than I expected. I spent an entire
day on code that I threw away, to try again the next day. So it goes.

I should be putting up another announcement in a day or so when the next
beta distribution is ready.
  -- Howard