There are solutions to offload IBM Z workloads. Most significant these days is probably Enterprise Server (ES), a former Micro Focus product that's now at Rocket. This is also interesting: it uses the (now former) Micro Focus COBOL and PL/I compilers plus scaffolding to let you move reasonably well-behaved custom z/OS applications to Windows or Linux.
The idea is that your average (if there is such a thing!) z/OS shop who wants to get off of z/OS moves some stuff easily: mail went long ago (and was probably on VM anyway), database, HR, payroll...all those have arguably reasonable alternatives. But then you get down to that huge core app that's been developed over the last 30/40/50/60 years, and porting that to Java or C or whatever ... well, you could do it, probably, but it would take years and cost millions of lives. And the big risk is that you miss some subtleties, and break things/lose $. The wise CIO/CTO at that point says "OK, what else can we do?" Well, with ES, you can probably/mostly move it unchanged. You recompile, and ES lets you use your existing JCL, emulates CICS, etc. You'll need to rework any assembler, of course. But many large applications really are pure COBOL and/or PL/I so that doesn't even come up, I'm told. Is such a port trivial? Still no, but the hard part--the application itself--needs little or no serious changes. Maybe some SQL changes for Db2 LUW or whatever ES supports, but that's relatively easy to test and prove. ES can run in EBCDIC or ASCII mode, too, so depending on the data, you can do what makes most sense for you. I've been fascinated by this for years, because it makes a LOT of money -- many companies use it -- yet I'd never heard of it, never see it discussed here or anywhere else. This isn't me saying "I think ES is great"; I really don't know. Again, I do think it's very interesting, and the fact that people DO use it successfully argues that it's not the fantasy that I sort of thought it was when I first heard about it. Anyway, your thing could perhaps do something similar, but I'd suggest that since these apps are probably more COBOL and PL/I than C, saying "You can recompile with gcc" isn't really going to solve most folks' problem. And of course there was also LzLabs, whom IBM basically killed earlier this year through legal action. Their approach was even more interesting, if scarier: realtime interpretation of IBM Z instructions on x86. Sounds like another fun project, but I'd be scared to run that in production, personally. Going back farther, there was PSI, who had an emulation layer built using the Itanium programmable layer (forget what that's called) that would do enough of z/Architecture that things could run. IBM bought and killed them (note that with LzLabs they skipped the "bought" step). You, at least, seem to be safe from IBM going after you. Microsoft? Not sure... -----Original Message----- From: IBM Mainframe Discussion List <[email protected]> On Behalf Of Paul Edwards Sent: Friday, October 24, 2025 9:53 PM To: [email protected] Subject: Re: Win64 x64 EBCDIC mini-clone On Fri, 24 Oct 2025 21:11:49 -0400, Phil Smith III <[email protected]> wrote: > Ok, that clarifies a bunch. Let me ask the hard question, > then: what is this for, other than a very interesting exercise? You tell me. You don't see any market for offloading processing from the mainframe to the PC? And not necessarily the PC - if you're happy to cross-compile, you could use an ARM processor. Including literally using a smartphone to do your EBCDIC data processing, especially with something like PdAndro (see pdos.org for link to that). I personally am not familiar with "the market" (I have more interesting things to do than market research). So I have no idea what possibilities may exist in the real world. I just saw a theoretical barrier to switching machines being a data conversion effort. You no longer need to convert your data in order to offload processing. And nor do you need to do emulation. You do need to write your application in a portable language though. But you can probably get around that too. Just not as elegant as C90. Nothing is. Including C99 etc. They did a really good job back then. BTW, the US government is giving away the standard for free: https://nvlpubs.nist.gov/nistpubs/Legacy/FIPS/fipspub160.pdf Including the Rationale. Called C89, not C90, then. But I was waiting with bated breath to see if ISO was going to change anything. Other than renumber sections, nothing was changed. You can still buy it too (or you could a couple of years ago anyway - I bought one), from Australian Standards. BFN. Paul. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
