Chris Craddock writes:
>>If they're written in C or C++, you recompile them and run them
>>on z/OS, probably with USS.
>Spoken like someone who's never actually tried it. There are things that
>port easily and things that don't. In my own (hands on) experience it
>isn't nearly as easy as the glossies would have you believe.
>Particularly with respect to code-pages, character sets etc.
Someone asked how you'd port from Linux to z/OS. I answered the question.
I think a fair reading of what I wrote would indicate that I'm not trying
to hide anything. C/C++ is not portable -- sorry to repeat that shocking
news :-) -- and z/OS is not Linux is not Microsoft Windows.
That said, for a BT/I with a new idea (which spawned this particular
discussion), would the BT/I be writing code in C/C++? Computerworld just
declared at least C a dead language, so does that mean it's official? :-)
>>If they're written in Java you just run them. No changes required.
>Again, mostly true, but not entirely. There are differing assumptions
>about the "shape" of the environment. Getting the JVM to run jar files
>originating from off platform is almost always entertaining.
If they're J2EE then the only thing you should have to worry about is
external resources. For example, if your J2EE application accesses an LDAP
user repository, then you'll want to re-point that access to, say, the z/OS
Security Server and its LDAP. You'll also have to populate LDAP.
If you're using JNI you're not writing Java, and you should know what
you're getting into already. :-)
On the scale of development difficulty (1 lowest, 10 highest), Java pushes
you down to 1.
>I have no idea about Perl. Getting Apache to run on z/OS was a far from
>trivial exercise. It can be done of course, but you end up having to
>make lots of changes to get it to run well.
Re: Apache, I had in mind the Apache HTTP Server mentioned in IBM
announcement letter 206-077 (WebSphere Application Server V6.1 for z/OS
announcement), though I know about the Apache for z/OS you're probably
thinking of (distributed via Apache.org). My expectation is that IBM's
packaging will take care of the "run well" concern, for a better "out of
the box" experience. That's not to criticize the long running/well
maintained distribution downloadable from Apache.org.
A lot of BT/Is are using the so-called lightweight "LAMP" stack
(Linux+Apache HTTP+MySQL+PHP) to build applications. This stack has a
different initialism on z/OS (zADP = z/OS+Apache+DB2+PHP), but from a LAMP
application development point of view it's very familiar. I know
internally we've been having some fun looking at running a "Wiki" on the
zADP stack.
I should also note that this constant stream of z/OS technology foundation
enhancements to ease application porting happens, as Alan said, through IBM
PWD. Software developers ask for stuff to make their jobs easier, IBM
listens and prioritizes -- yes, really -- and stuff happens. The recent
list is even longer than I've described, and there's more coming (such as
z/OS 1.9 in September as mentioned). Want more? Is there more technology
that would make it easier/quicker/cheaper/faster to bring your applications
to z/OS? Ask IBM PWD.
Re: Mainframe access, there are these EAL5-certifiable things called LPARs
which every modern mainframe has. They didn't exist in the 1970s. Anyone
working at one company (or solo) accessing another company's development
z/OS LPAR remotely? This is common from what I can tell and getting more
common.
Wayne Driscoll writes:
>No offense, but if I am
>going to write a linux product, I can spend less than a thousand dollars
>US and get a dual core intel box and put Fedora linux on it, and I can
>start writing it. Why would I give a hoot about running it on a z box...
I thought this was common knowledge (how to create Linux on z application
code), but let me expand on that. It's a good question.
There are at least two routes. One is that anybody with a dream can get a
free Linux z/VM guest (including root access) for a period of time from
IBM. That's the "Linux Test Drive":
http://www.ibm.com/servers/enable/site/testdrive/zseries/
But you don't actually even need that except perhaps for final testing.
You can compile Linux on z binaries on practically any system that can run
Linux, including Intel/AMD (i.e. your desktop PC). Linux is Linux. This
is pretty well documented and is called cross-compilation. For example, if
you're talking about C/C++ code, then you'd be using gcc and simply add the
-ms390x (for 64-bit) or -ms390 (for 31-bit) switch to your command line.
You would also need the applicable s390x or s390 libraries copied to your
Linux drive so that gcc can access them during compilation. These binary
libraries are available with all the mainframe Linux distributions. See my
previous note for where to download those distributions.
This is one of the reasons why you see so many binaries for s390 and s390x
available for download from Linux software project Web sites.
I posed the question way earlier whether Linux on z is one significant
on-ramp for the BT/I who ultimately wants to build a new company and (also)
ship their new product(s) on z/OS. I would argue yes since I'm already
seeing that pattern, but perhaps opinions differ. (I also think it
shouldn't be the only on-ramp, and it isn't.) By the way, since the
primary way to develop for z/TPF is by using Linux, Linux had better be an
on-ramp for z/TPF. :-)
Howard Brazee writes:
>Our infrastructure is a lot
>different nowadays - you can have a share of a mainframe on the other
>side of the world and you don't need a power strip for it.
A very astute observation -- I totally agree. A huge chunk of India's
economy depends on this new reality, for example.
Mark Zelden writes:
>An *IBM representative* used the "U" word not once, but twice.
As mentioned countless times, and which I guess I have to repeat yet again,
I'm not an IBM representative here in IBM-MAIN. I speak only for myself
here. Didn't you also see the word "orgy"? I don't think that word is in
the IBM PR handbook. Does IBM have a PR handbook?
Ray Mullins writes:
>Random thought - I wonder what would happen if Fujitsu and Hitachi decided
>to release their clones of MVS and VSE to hobbyists. Yeah, yeah, there's
>legal agreements, etc., which probably preclude that.
It's an interesting hypothetical to ponder for a little while. But I'll
ask a question that'll give you a hint what I think the impact would be.
Have these companies shipped any substantial innovations or improvements in
their operating systems in, say, the past decade?
Dave Kopischke writes:
>http://publib.boulder.ibm.com/tividd/td/TDS390/SH19-4494-
>06/en_US/HTML/drlb9mst389.htm
Sorry to burst anyone's bubble here, but this is a common typo. To prove
the point, use Google to search on "Linux on z/OS" (with the quotation
marks). You'll get way too many hits, including some embarrassing ones at
ibm.com addresses. If this is some new super secret z/OS subsystem, IBM
isn't doing a very good job keeping it secret, and why would it need
Hipersockets to communicate with itself? :-)
I should start putting "Linux on z/VSE" in random blog posts to see if I
can start another rumor, just for fun. Ooops. I guess that just went into
Google now.
- - - - -
Timothy Sipples
IBM Consulting Enterprise Software Architect
Specializing in Software Architectures Related to System z
Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific
E-Mail: [EMAIL PROTECTED]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
