Wratko Hlavina

If I may offer an opinion: * HTTP is fundamentally streamed I/O, whether you send the content-length _a priori_, or you used chunked transfer encoding (which you may do if...

To compare with industry job schedulers: * [Grid Engine/SGE/UGE](https://docs.oracle.com/cd/E19957-01/820-0699/i1003714/index.html): Resource requests are minimums, may be hard vs soft, and are specified in order of importance during job submission. Hard resource...

I agree that just because many batch schedulers do a thing, does not mean a workflow description should do that thing. Rather, I was illustrating the current landscape to benefit...

I agree too, that you're solving a different problem. The use case for the job schedulers, is one where a given compute cluster is concrete and has known resources available,...

Perhaps expanding the scope a bit regarding conformance tests, our experience finds that runners have problems with edge cases in path mapping, and these are most pronounced when combining the...

P.S. The bug which affects the reference `cwltool`, common-workflow-language/cwltool#1029, is causing us some issues. If that one could be prioritized higher, it would help us out. I've looked at the...

```workreuse```, from #583 as I linked, is somewhat related, but I'm not sure it fully addresses the use case here. This one is more about having a debug mode, for...

The pragmatic constraint is that in non-trivial workflows, there will be large intermediates that need to be passed between 2 tasks, and that implies they must be outputs (not in...

Clarifying: A runner may be able to automatically make some judgement calls based on space/time/compute calculations, but the feature I am requesting here regarding the CWL specification, is formalizing a...

Thanks for the feedback and guidance for implementation. I may not have time to attempt coding a proof of concept at this time, but I may give it a shot...