Manifest of MATLAB image to be restored

[vijayiyer05]

Add description of MATLAB image to be restored (with small refinements) to working order in the course of addressing #276.

As discussed w/ the DANDI team, it's envisioned to share/track the contents of MATLAB (and potentially other) images on this or a README. This is a starting point to be used concurrent with #276.

The location/format/etc for this and other DandiHub image manifests can be updated later.

[satra]

thanks @vijayiyer05 - we'll keep this open and till the other issue is addressed.

[asmacdo]

@aranega could you have a look at this, I'm not sure how to move this one forward. Do we need to do any pinning in the Dockerfile to make this true, or maybe we should update this list to reflect what is in the Dockerfile?

[aranega]

@asmacdo I think this PR is mainly outdated. The main information we needed to follow inside, i.e: which release (automatic latest release, or latest commit) is implemented already in the Dockerfile, and I think the versions that @vijayiyer05 pinned are indication of the version that is pulled in the image more than version constraints, but I might be mistaking.

@vijayiyer05 Currently, the version that are fetched when the image is built from the Dockerfile matches the version that you set in the README. With the current mechanism, if an update of Brain-Observatory-Toolbox or DeepInterpolation is pushed on github, this latest version will be picked automatically when the image is rebuilt. Should we pin exactly the version to the one written in the README or fetching the last version automatically as it's currenlty done is preferable for those dependencies?

[vijayiyer05]

Thank you @asmacdo for resurfacing this. Thank you @aranega for commenting.

This PR is really about establishing a placeholder for where the manifest might live in a non-Dockerese human-readable form, most likely in MD format as a table as I set up for this PR. This is important to explain to new/casual/deciding stakeholders quickly what's on the MATLAB image. We need this before we can set up the maintenance project series.

Given that things have stabilized and there is a MATLAB subfolder in the images folder, I suggest the table contents could be moved to a new README file I could propose there.

Would this be agreeable to all here? If so, I can make it so in but maybenot till week of 3/29.

Re my testing of the live image against that manifest:

• Line item 13 seems missing (the Deep Learning Toolbox Converter for Tensorflow support package)
• Suggest to leave BOT & DeepInterpolation as is, though yes I suspect release anchoring will make sense at our next maintenance tic

[aranega]

@vijayiyer05 Thanks for the feedback!

Regarding the Deep Learning Toolbox Converter for Tensorflow, the package is installed (installed here in the Dockerfile ). The fact that it doesn't appear might suggest that there is an extra action that needs to be done to enable it?

No problem for anchoring specific version, we can discuss that in our next maintenance sow, we have all the framework for it, so it will be just a matter of deciding what we want to do, on a technical point of view, it shoud be trivial :)

Regarding the update of the PR, I can take care of it if next week if that makes it more convenient for you guys.

[vijayiyer05]

Thanks @aranega for noting the support package wasn't overlooked in the Dockerfile. I will take a look next week to see if another downstream step may have been omitted from the spec.

Sounds good on other points (anchoring at next maintenance timepoint, updating this PR to propose a new dedicated README w/ the manifest).

[aranega]

@vijayiyer05 I created a new PR #290 to deal with the new README. I refreshed some sections at the same time, adding some new information about the Latest Commit entries, double-check/updated the versions of each lib if necessary, and refreshed the end of the README that was a little bit outdated.

I couldn't push on your repo or update this PR, so that was the fatest option. If you prefer that I open a PR against your repository instead, let me know :)

[vijayiyer05]

Superseded by #290.