-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathyoh-notes
54 lines (37 loc) · 2.88 KB
/
yoh-notes
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
preparing an image-- running out of space
- podman vs docker
- approaches we didnt take
- piptimemachine
problems
- no cleanup
remaining problems with the current approach
install reproman on discovery
2023-10-05 -- some notes. more highlighted in local pdf, heh
- hard to read since often dwells on something only loosely related to the point to deliver or just sounding "sophisticated"
- "lessons learned" is formulated as a dry " We document a number of prominent difficulties with de novo article gen‐
eration, arising from the rapid evolution of extrinsic tools, and from nondeterministic
data analysis procedures. "
Why do you think people should read the article? what should they expect to discover/learn after reading abstract?
IMHO this should be the main theme for the paper, not just a side note: We
have tried to reexecute claimed to be a reexecutable publication, it was
supposed to be a quick easy task. We ended up discovering many nuisances which
we think worth telling people about. We also tried to minimize changes
ecessary to be done to original paper "setup".
- when talking about your own work, it should be made more clear and not to make superficial claims such as
"emerging infrastructure standard, RepSeP [5],"
- "silo'ed introduction": introduction tries to guide between degrees of reproducibility/reexecutability but makes not a single reference to existing overviews and provides no pointers to existing other efforts.
- e.g. there is https://elifesciences.org/for-the-press/eb096af1/elife-launches-executable-research-articles-for-publishing-computationally-reproducible-results , https://neurolibre.org/ and so on -- many other efforts and formalizations
- when introducing OPFVTA article doesn't provide succinct summary , e.g. at least to mentioned that it was mouse whole-brain opto-fMRI study
- "Errors should be fatal more often than not." -- it is just about default behavior of shell scripts... has little to do with make or anything else!
also IIRC we had a case where it was some try/except swallowing some errors... or am I dreaming?
Chr:
There was a case of `|| exit 0` to mitigate pdfdiff exit codes for differing documents, we now simply have an echo statement.
Very niche concern which I don't think any broader lessons can be based on.
- Best practices -- I think would benefit from concentrating on what matters and may be giving an example?
also comes from YODA. Also allows to minimize number of bind mounts needed for container runtime specification
- Re Containers ... small -- mention hint on bind mounting "adjusted content"
from the host system while working quinks out or even persistently.
- re debug outputs. Also recommend to just use stdout/stderr to communicate
- for freezing environment, mention other possible solutions such as use of
APT snapshots and nd_freeze here, can even reference that neurodocker has
an option for it.