RFC: Reinvigorating the Haskell Symposium

[jgbm]

This is an RFC regarding planned changes to the Haskell Symposium: inviting new types of contribution, and submitting accepted papers to JFP rather than having a proceedings.

Rendered.

[simonpj]

There is a fork in the road here.

• EITHER (A): continue with the Haskell Symposium as an archival ACM publication, with full reviews etc. Consequence: a HS paper is a publication and you can't resubmit that paper elsewhere.
• OR (B): move back to a more informal workshop format, in which there is no published proceedings, and an accepted paper or talk can certainly be submitted elsewhere for publication.

We have been doing (A) for some time, but the steering committee is signalling that (A) isn't working very well; in particular we aren't getting enough submissions. A contributory factor could be (although it's hard to be certain) that conference ranking systems (e.g. http://www.conferenceranks.com/) don't rate the HS highly, which makes authors think about submitting elsewhere.

Hence the suggestion to move towards (B), while still providing authors with a route (but not an obligation, nor a guarantee) of an archival publication in JFP.

@ivanperez-keera points out that it is much easier to get your institution to fund a trip to ICFP if you have a publication at ICFP. True enough. But we also need to balance the fact that the majority of participants are not authors of accepted papers; and that the declining submission rate must be telling us something about the attractiveness of HS as an archival venue. Just doing more of the same is possible, but personally I think we might want to be more proactive than that.

The proposal is for consultation; it's not decided! If you have better ideas, please do offer them here.

[goldfirere]

We could have our cake and eat it, too.

(Aside: I'm in strong support of having talk proposals as a separate category. I want to hear more cool ideas that are cooking in Haskell-land, even if they're not yet polished and are out-of-scope for HIW. There seems to be no controversy on this point.)

The interesting idea presented here is linking HS presentations with JFP. There is some debate, however, about the timing of the relationship. Why not have this aspect of the relationship be flexible? That is, authors who want to present their latest work at HS can submit to HS, get a review, present, and then post-submit to JFP. In addition, authors can submit to JFP directly, get their paper accepted, and then present at HS.

One question: why would a JFP author choose to present at HS rather than ICFP (who also accept JFP presentations)? This is a bit harder. We could simply open up this option and see if anyone takes us up. Perhaps someone would if they're more interested in having their travel budget include HS than ICFP. Or maybe ICFP has to be selective in inviting JFP authors who want to present -- we should ask. If they are selective, we can offer HS as a second venue for presentation. Or we could make the HS PC serve an entire year, and then JFP submitters can opt to be reviewed by members of the HS PC for their initial review, even if off-cycle. A paper accepted by this route would be presented at HS, not ICFP.

High-level piece: thanks @jgbm for thinking creatively about how to evolve the HS. I think this kind of innovation is needed.

[ulysses4ever]

There is a fork in the road here.

Not necessarily. Some conferences have been employing a mixed model for some time. ECOOP, for example, has what they call Journal After with a similar strategy to what’s proposed here (there are three journals they collaborate with). If you go that route, ECOOP will still publish an extended abstract of your paper. (And you get to present it, of course.) But that’s optional: the “normal” publications are still the default option.

[jmct]

In recent years, submissions to the Haskell Symposium have dropped sharply, making it increasingly difficult to compile a compelling event.

A compelling two day event, or a compelling event, full stop?

Perhaps Haskell has reached a state of maturity where there isn't the flurry of research about Haskell that there once was, and a single day event (like ML-Family, Scheme, etc) is more appropriate?

I also like the point that @ulysses4ever brought up, a mixed-mode approach could be worthwhile, if we think there's enough for a good single-day event.

If the issue is that there are too few submissions for even a single day event, it might point to different problems.

Aside from that, I also feel a bit sad that bibliometrics are a factor here, but I recognize that I have the privilege of not worrying about that in the way that others must. I

[ivanperez-keera]

That is, authors who want to present their latest work at HS can submit to HS, get a review, present, and then post-submit to JFP.

@goldfirere If the paper does not appear in any proceedings, for this to count as a publication, authors will have to go to the Haskell symposium and also pay JFP which IIRC has its own fees. That would act as a sieve that would prevent people who can't afford it from publishing otherwise great work.

[adamgundry]

I wonder about the relationship between this idea and the future of the Haskell Implementors' Workshop. HIW has typically been a venue for more informal talk proposals, but I believe it has also seen a decline in the number of submissions. Moreover my impression is that ZuriHac now attracts more of the "open-source Haskell hacker" crowd than ICFP does.

Perhaps rather than having both HS and HIW co-located with ICFP over 3 days, it would make sense to have a single 1-2 day Haskell event at ICFP, and a HIW-like event alongside ZuriHac (somewhat similar to the GHC Contributors' Workshop last year)?

[goldfirere]

Good point @ivanperez-keera -- we don't want to increase costs.

It might be interesting to experiment co-locating HS with ZuriHac one year. (I think the ICFP steering committee might be happy to co-locate all of ICFP with ZuriHac at some point, if ZuriHac can move to the fall for the year.) I think expanding the talk options around ZuriHac is a great idea. The only drawback is that its location is fixed, meaning that some people are always going to have a hard time getting there.

We might be able to do both here, too: have one day of HS co-located with ZuriHac, and have the other day co-located with ICFP. Authors then have a choice about which venue is better for them (including timing, location, audience).

[ivanperez-keera]

My interpretation of what @adamgundry was saying, and correct me please if I'm wrong (I don't want to speak for anyone), is that the HIW aligns more with Zurihac in terms of style and topics, so maybe something like that could be co-located with Zurihac, and something like the HS could be co-located with ICFP and cover part of the spot left by the HIW.

[jgbm]

I think the arguments for continuing to have an archival proceedings of the Haskell Symposium are compelling (for now).

The updated proposal allows 3 kinds of submissions:

1. Talk proposals
2. Full papers, for publication in HS
3. Full papers, for publication in JFP

Everything accepted will be presented at HS. Accepted full papers in category 2 will be included in the proceedings of the Haskell Symposium. Accepted full papers in category 3 will continue to JFP, as proposed before.

@goldfirere raises a challenging point: given that JFP articles can appear in ICFP, why would an author who was convinced that their paper was JFP-ready prefer to present at the Haskell Symposium?

One answer I can imagine is timing: for work that just misses the ICFP deadline, it might be preferable to present the work sooner, while still having credit for publication in a higher-tier venue.

There have been several suggestions about relocating either part of the HS, the HIW, combining with related events, and the like. I think these are interesting suggestions, but I think there would be value in seeing how the number of submissions changes with these proposals before proposing further (and more dramatic) changes.

@adamgundry points out that this could lead to HIW talks being sent to HS instead. I hope this is not the case. Looking at last year's HIW talks, many of them focus specifics of implementation that might not be suited to HS. In the longer term, I can see it making sense to reduce HS to one day, and have HIW as the second day of Haskell-related content at ICFP.

[simonpj]

@jgbm proposals sounds sensible to me.

[sgraf812]

@goldfirere raises a challenging point: given that JFP articles can appear in ICFP, why would an author who was convinced that their paper was JFP-ready prefer to present at the Haskell Symposium?

From what I gather, it's not that the JFP article "appears in ICFP" in the sense of an ICFP proceedings publication; it's just a talk slot. I had a brief look at the titles of all JFP talks from ICFP 2020 to ICFP 2023; the only talk that was seemed specific to Haskell was Modal FRP for all: Functional reactive programming without space leaks in Haskell. I think the risk of cannibalisation is reasonably low.

[jmct]

If @jgbm feels that the Steering Committee has gotten what it wants out of this RFC, we propose archiving the discussion. We're going to create a process for archiving RFCs that will make it easy to review the discussion, while closing the PR itself.

However, if you feel that there's more to discuss, we can delay archiving it and try to get more eyes on the RFC.

[jgbm]

@jmct I think archiving this discussions makes good sense.