IRC channel logs


back to list of logs

<civodul>i'm reading about workflows and reproducibility
<civodul>it's a real bummer that we don't have an article about GWL
<civodul>or even a blog post
<rekado>I’m a little too shy when it comes to the GWL
<rekado>it’s a few more hacks away from being fast enough for real workflows
<civodul>heh i see, performance seems to be a problem
<rekado>5 months ago I added workflow-files and workflow-hash, which was supposed to be the first few steps towards more aggressive caching.
<rekado>re blog post: we have this:
<civodul>right! the book chapter should satisfy folks in academia
<rekado>(it’s rather outdated…)
<rekado>but the core idea hasn’t changed
<rekado>when the cold weather has passed and I’m less likely to be sick every two weeks I’d like to finish up a few features and write a new paper on the GWL
<zimoun>civodul: as discussed with rekado, I plan to write something similar as short article about Guix in Nature Scientific Data for GWL.
<zimoun>Indeed, it misses some bits to be fully usable but I do not think it should be a blocker for communicating. :-)
<zimoun>By “I plan”, I mean more or less this year. :-)
<zimoun>One key point about workflows is that they always rely on infrastructures (supply chain) that are weakly reliable when the time is passing.
<zimoun>it could be one key even if GWL is slow or missing some features if it has the ability to just work, running a pipeline of 3 years ago.
<zimoun>When speaking about passing of time, there is our approach (freeze) and the other approach (make tools evolving).
<zimoun>For an example of the latter, see for Machine Learning.
<civodul>zimoun: nice!
<civodul>i think we need to clarify our position
<civodul>it's kinda weird, i feel we've been repeating the same things since the 2015 paper
<civodul>yet it seems we're missing something to point to
<zimoun>well, we need to say again and again and differently the same thing in order to spread that very same thing, I guess. :-)
<zimoun>about GWL, I agree that something to point to is missing. Roel’s master thesis is somewhere in the mailing archive. :-)
<rekado>oh, Roel’s master thesis mentions the GWL?
<civodul>ACTION didn't know about the master thesis
<rekado>there’s also the upcoming workflow engine comparison paper; once that’s out I’ll add it to the publications page.
<rekado>zimoun: I’m all up for writing that paper with you (if I stay healthy for long enough)
<rekado>would like to point to a real workflow as an example, though
<rekado>I once translated the pigx rnaseq workflow, but it doesn’t really do anything to highlight unique GWL features
<rekado>(other than those that the GWL inherits from Guix)
<civodul>rekado: the comparison paper is the one you mention in the activity report, right?
<zimoun>My holidays are next week (if the French strike lets me go far :-)). Well, I will try to open a thread on gwl-devel to exchange about what we think as unique GWL features.
<civodul>or not you, your colleague
<rekado>civodul: yes, that’s the one
<rekado>at the time we only had a preprint
<civodul>right, that one looks super interesting
<zimoun>that’s not their Master thesis. But they sent me privately.
<civodul>now that rings a bell
<zimoun>Well, it is CC-by-SA so I guess we could put it in GWL website.
<rekado>neat, I’ll add it later
<zimoun>Do I need to ask to Roel if we can publish on the webiste their Master thesis? Because I have this thesis from a private email.
<civodul>i think so