[Xitami] Multi server issue

Paul Reid preid at rci.rutgers.edu
Wed Mar 26 06:48:40 CET 2008


Here's some sloppy thinking from someone who knew the ARPAnet, drifted 
away, and on return had to learn this newfangled TCP stuff.

 > You have one port 80

IP is street address.

Port is a box at that address.

Clearly you can have 1234 Oak St Box 80, and 1235 Oak St Box 80, without 
mix-up. 192.168.1.100:80 and 192.168.1.101:80 are -different- logical 
places, even if they are same physical card in same machine.

What RO was trying -SHOULD- work.

Well, maybe. I'm constantly getting mail for "141" in other 
buildings/streets. My mailman does not always get it right.

Hierarchical addressing systems are easy to screw-up. Mailman sorts 
right only 99% of the time. If you are building a machine, it is easy to 
overlook something, get it 99% right. If it screws-up only in unusual 
situations, it may go unnoticed for a decade.

If this were Mac O/S 7, or Win 311, I'd bet on the tossed-together 
non-unix TCP stacks of that vintage. But W2K+ stole a unix stack and 
should do the right thing.

Magnus has magnanimously traced the conflict. Xitami opens a lot more 
ports than just 80. My reading is that some of these are opened without 
specifying an IP. The default is "all", the first copy already snagged 
those ports, the request is refused, Xitami suspends itself.

Is it a bug? Since you CAN run MANY different IP addresses on a single 
copy of Xitami, demanding to be able to run two copies seems greedy. 
(Specially for free.)

Yet clearly there is IP:Port support; just somehow not on the LRWP 
frill. Probably an oversight by the developer. Well, it IS a technology 
demo, not promised to be all things to all people. And it appears the 
community didn't notice this oddity for many years; how "essential" can 
it be?

Xitami NT-Service source code IS available. And it is darn clean stuff. 
Knowing the bug location, any half-fast C programmer should be able to 
find the function call and the oversight. (Or, since "nobody" uses LRWP, 
simply lose the whole LRWP call.)

Nearly a decade ago, I was able to build Xitami from source. I'm not so 
clever anymore, or I've grown smart enough to have lost all my MS-C 
compiler disks.  I'm a user, not a coder. Do we have any coders here, or 
just users? And coders who want to spend time on this odd-horse problem?

I think the xitami.wikidot.com site says it all. There are NO edits, 
ever, not even the glaring typo in Pieter's introduction ("not" for 
"now"). I just fixed that, and added Magnus's comments.... but 45 days 
with __NO__ community input means nobody has an oar in the water.

  -PRR


More information about the Xitami mailing list