On 8/20/2010 1:14 PM, DLee@xxxxxxxx wrote:
1) Should programs that are called thru the client URL be in a special
library, or can they just continue to reside in the normal library list?
To simplify security, it might be a good idea to put your programs that
are accessible via HTTP in a separate library from those that aren't.
If you have multiple libraries for traditional programs, then it makes
sense to have multiple libraries for your web programs as well. (i.e.
have MYLIB and MYLIBWEB, then MYLIB2 and MYLIB2WEB, etc.)
Just my opinion.
2)I understand that two properties make a server instance unique, the host
name, and the port. We only have on iseries(host),
I don't understand why people run multiple web instances in production
(it makes sense in development, so each developer can have a separate
sandbox... but in prod, there should only be one instance, IMHO)
What makes a unique server instance is that there's a separate copy of
Apache running with a separate configuration file, etc. It's true that
incoming connections decide which instance they are connecting to by
choosing which IP address and port they connect to.
You can also have virtual hosts -- which means a single Apache instance
behaves differently depending on the domain name, IP address, or port
number used. In that case, there's only one copy of Apache running, and
one config file, but possibly multiple ports, domain names or IP addresses.
So I guess we will be depending on the port to make server instances
unique. Correct?
I guess.
3) How are instances usually setup, according to applications like
accounting applications, insurance applications etc?
I'd run all of these on the same instance. What's the value of running
them on separate instances?!
Honestly, before Apache came to IBM i, I'd never even heard of a shop
running multiple instances simultaneously. Lots and lots of people
using virtual hosting, though... that's very common... but not
instances. I always knew you could do it -- run two copies of Apache
at once... but never saw a reason to do it, and never heard of anyone
else doing it. Til IBM brought it to IBM i, and everyone on this
platform seems to do it.
Personally, I think it's silly. Why use extra resources? Why require
multiple ports to be available through firewalls? (Including
nonstandard ones.) What do you tell folks who can't just communicate
on any arbitrary port, but need to use your services?
4) Would appreciate knowing about a manual that might answer some of
these questions.
Me too.
5) Would appreciate knowing some questions I should have asked.
Me too.
midrange.com
