[Info-vax] Using VMS for a web server

[David Froble]

JF Mezei wrote:
> Suggestion to VSI on Apache:
> 
> It is unclear to me how much patching is required to make Apache run on VMS.
> 
> Ideally, customers should have a "kit" to which they can apply changed
> Apache stuff and rebuilt the VMS version.
> 
> This would give VMS customers "instant" access to patches and new
> features while allowing VSI to focus only on the modules that are very
> VMS specific.
> 
> For the "standard" modules that have only minor modifs in the code, VSI
> should embed them with #ifdef stuff in the open sourced code base once
> so they don't have to constant do it.
> 
> 

I know nothing, but if I did know something, one thing might be that one 
of the really big issues is FORK.  VMS doesn't FORK, and I don't miss 
it.  Unfortunately, VMS doesn't make it easy to work around, with 
respect to accepting a socket connection and having each processed while 
additional connection requests are serviced.  The documentation on this 
in the TCP/IP docs is non-existent.

So it's most likely not some little things in Apache, but at least one 
BIG thing.  The way I understand the VMS workaround in Apache is a set 
of sub-processes is set up and used as required.  Not sure what happens 
if that set is all in use and more connection requests come in.  Maybe 
it gets expanded.

I'm not even sure it should be sub-processes.  Once a task is assigned, 
it should finish, and not be terminated if the parent process has a 
problem.  At least that's my plan, if I ever get back to that task.

Unix is not VMS, and VMS is not Unix.  Expecting one to work like the 
other just isn't reasonable, in my opinion.  That leads to Unix 
applications perhaps not working so well on VMS.  To me the problem is 
not being able to port Unix applications to VMS.  The problem is not 
having the functionality in VMS applications.