(review) embed much more metadata in talks released to video.fosdem.org

[markvdb]

Sreview does not preserve the pretalx talk id, and only limited metadata inside the video files it creates.

Embedding richer metadata could really make wider reuse outside video.fosdem.org easier. Examples that come to mind:

• German and UK university staff asking to add DOIs to FOSDEM talks (private requests)
• Newpipe requests (see also FOSDEM Videos TeamNewPipe/NewPipe#5551 )
• Peertube hosting (see also video torrents/peertube #79 )
• ...

In the mean time, pretalx (and for past editions pentabarf) metadata could really help. See https://fosdem.org/2026/schedule/xml and relevant past schedules according with similar naming schemes, example https://archive.fosdem.org/2025/schedule/xml .

[DamonHD]

Providing DOIs for the talk videos and/or textual summaries under fosdem.org makes them easier to cite from formal academic papers, and indeed lends them a little extra credibility, in part because of implied longevity and canonicality, ie DOIs are scientific canonical permalinks.

This would also be useful for each FOSDEM event as a whole, and (say) each track/devroom, and should be relatively easy to accomplish mechanically.

Using the pretalx/pentabarf metadata to provide filled out citations such as BibTeX would be extra value, such as with this gem:

curl -sSgLH 'Accept: application/x-bibtex' "http://dx.doi.org/$1"

[yoe]

Sreview does not preserve the pretalx talk id, and only limited metadata inside the video files it creates.

Detail, but pet peeve: SReview or sreview, please. Not Sreview :-)

I guess 'limited' is in the eye of the beholder; we do set quite a number of metadata items. Exact code here:

https://github.com/yoe/SReview/blob/8c58c9b42a6e52b2354810ae23c63f5edfa7c073/scripts/sreview-transcode#L356-L379

We set container metadata for track, room, date, title, subtitle description, speakers, and the URL on the FOSDEM.org schedule. You can see this metadata if you do things like ffprobe on the file; some media players will also show it when you open the file. I'm not sure if the configuration item for the video license is correctly set; if it is, then that will also show up.

Note that which of these items actually end up as container metadata depends on the container in use. WebM/matroska allows for custom metadata items and therefore allows for basically anything; other containers are more limited. I am not sure how much MP4 allows, it may not show everything.

Some of this metadata is also added to the opening slide. More can be added there, but I also don't want to crowd that slide.

Adding the pretalx id is reasonably trivial to do, as it is already stored in the database. The hardest part here is coming up with a name for the field in the container metadata, remembering that SReview is used by more than just FOSDEM, so just using the name 'pretalx id' or some such would not be the correct way to handle it. I could add it as 'talk id', or make a configuration item.

For the DOI reference, I would need to know more about how they are formed.

If we register a namespace and are then able to generate our own DOI references within that namespace, then that's something SReview could handle. If more work is involved, then I would suggest that some other process handles this, and that the DOI reference is added to the website schedule page.

Either way. SReview could certainly look at posting bibtex information some place if someone wants to provide a perl snippet that generates it correctly from the SReview database schema.

[yoe]

I'm not sure if the configuration item for the video license is correctly set; if it is, then that will also show up.

It turns out that yes, it is. Running ffprobe on one video shows this, amongst others:

Input #0, matroska,webm, from 'UHDG9Z-who_funds_the_egg_cracking_the_foss_funding_paradox.av1.webm':
  Metadata:
    title           : Who Funds the Egg?
    SUBTITLE        : Cracking the FOSS Funding Paradox
    SPEAKERS        : Deborah Udoh
    EVENT           : FOSDEM 2026
    RECORDING_LOCATION: Janson
    SYNOPSIS        : <p>Every successful open source project starts small, but not every small project gets the chance to succeed. Here’s the paradox: funders prefer to back proven impact, yet impact requires early support. If new initiatives can’t access resources until
                    : <p>This talk explores this chicken-and-egg dilemma and proposes ways to flip the script. What would it look like if we invested not only in impact already proven, but also in potential?</p>
                    : <p>Drawing from my experience building <a href="https://www.linkedin.com/company/pre-seeds-research-101/">Pre-Seeds: Research 101</a>, I’ll share insights into the struggles early-stage projects face (e.g., limited visibility, lack of credibility, and d
                    : <p>Attendees will leave with concrete ideas for how they — as individuals, organisations, or communities — can sustain the pipeline of emerging projects, ensuring the long-term resilience of open source.</p>
                    : <p>If we want an open, thriving, and continuously renewing FOSS ecosystem, we need to get better at nurturing the eggs, not just celebrating the chickens.</p>
    DATE            : 2026-02-01 11:30:00+01
    URL             : https://video.fosdem.org/2026/janson/UHDG9Z-who_funds_the_egg_cracking_the_foss_funding_paradox.av1.webm
    LICENSE         : CC BY

As said, adding more is certainly doable.

[yoe]

So I found the DOI handbook at https://www.doi.org/doi-handbook/DOI_Handbook_Final.pdf

If we can be pointed to a registration authority which allows us to register an organisation prefix, then I would suggest we register one of those and then use a 'year/id' format, with 'id' being either the Penta or pretalx id. We can do this retroactively for past years, although I do know that some other people have registered DOI references for at least some of our videos in the past; I didn't read the handbook in full (it's long), I don't know if it's allowed to have multiple registrations for the same resource like that.

[DamonHD]

DOIs are intended to be unique, but in practice they are sometimes not.

If you provided a comprehensive back catalogue of DOIs for all FOSDEM talks that you can, linking to the video and the synopisis and track and conference, my guess is that most observers would see that as a huge positive. Overlapping with some existing DOIs would not negate that.

Rgds

Damon

PS. Here is a DOI for an old FOSDEM talk in case that helps: 10.5446/32595

[yoe]

If you provided a comprehensive back catalogue of DOIs for all FOSDEM talks that you can, linking to the video and the synopisis and track and conference, my guess is that most observers would see that as a huge positive. Overlapping with some existing DOIs would not negate that.

That's great to know, thanks.

I committed a patch to SReview which makes the set of metadata written to the output files configurable with the values being templates that can be designed however we want. If a DOI naming scheme was set up such that the year and pretalx id were used as the reference, then we could easily create a template which uses that format.

I haven't tested this yet, so this is something for next year, not this year.

I also found datacite.org, a DOI registration authority which provides a REST API and which also lets you choose your DOI format yourself if you want. This might be a good candidate for registering our DOIs, though I couldn't find anything regarding pricing or the likes.

[yoe]

though I couldn't find anything regarding pricing or the likes.

So I had a bit of a closer look now, and found https://datacite.org/fee-model/

Given that, based on our status as a non profit and an estimated 1000 DOIs per year, we could expect a yearly cost of €500 (membership fee for non profits) + €800 (1000 DOIs @ €0.80 per DOI in the lowest tier) = €1300. We'll have to decide whether this is something FOSDEM should pay for at the next meeting.

However, if we also want to provide a back catalogue of DOIs for all the shaft existing videos from previous years, then we'd likely be paying the fixed fees of €2600 (membership plus a fixed amount of 2100 for up to 10k DOIs) for two years, as your tier is determined based on the amount of DOIs you registered the previous year.

I'll have to contact them to see if a social arrangement is possible where we do pay for the higher tier one year but then go to the lower tier the next in spite of the high amount of registered DOI numbers.

[DamonHD]

FWIW I normally donate ~EUR50 when I turn up in person, but was not there this year (though obvs I used the video streams, so I am some sort of idiot!), so please regard this as a small pledge towards costs in case that helps...

[saerdnaer]

Yoou can also ask libraries if they can provide DOIs for your recordings e.g. TIB already imports FOSDEM recordings via schedule.xml, and creates DOIs for them: e.g. https://av.tib.eu/media/41309 has DOI https://doi.org/10.5446/41309

They can also provide an API route which exposed all the DOIs for the FOSDEM talks they already imported on their "AV Portal".