I may be naive here, but isn’t it just a bunch of Perl scripts wrapping the aligners and handling the conversion? Replace the Perl wrapping with Nextflow, and replace the conversion with either the Perl sections or something faster. Thus we pull the logic out of the Perl scripts and into Nextflow natively, so we can handle things like how the fasta gets passed to the aligners.

Kinda like https://github.com/complextissue/pytximport R -> Python Perl -> Nextflow

#C073GFZ68PR you go

For a simpler example, stacks has a batch mode which would be much better in Nextflow, but is treated as one piece of software. So at what point do we say it’s one software versus a complicated script? Seems very opinionated to me. • how much work is it to rewrite in Nextflow? • what are the benefits or disadvantages? Can you measure them? • what is the long term maintenance strategy?

In 20 years I'll tell the next generation of bioinformagicians "Look at this Nextflow script I wrote during my PhD!" like my PI does to me with Perl scripts.

"Just a wrapper" is a bit of an understatement, it's many thousands of lines of battle tested code

I don't think that the Perl language is a bottle neck for speed, if it's slower than other tools my experience is that this is because it's doing more stuff