How do journals serve the TCS community?

Question

In the past, journals were the main way scientific/mathematical discoveries were disseminated and vetted. In some areas, they still are. However, in (theoretical) computer science that role is almost entirely performed by conferences and open web-based dissemination (e.g. arxiv or personal homepages).

There are still TCS journals (like ToC and JACM), but they seem to primarily publish papers that have already appeared in conferences and usually have an arxiv preprint. So I don't understand what value they are adding.

What purpose do journals serve in TCS?

I have heard three arguments in favour of journals, but I don't find any of them convincing:

1. Journals supposedly have higher standards for reviewing than conferences. It is true that conference reviewing is under greater time pressure, but it is still the same pool of reviewers (and the amount of time the reviewer is given to read the paper has very little correlation with how long they actually spend reading the paper). As a reviewer, I don't fundamentally treat a journal review request any differently than I treat an external review for a conference.
2. Journal versions are higher quality than conference versions. This is true, but it is moot, as most people will read neither. When I want to read a paper, I search for it and click on the most promising link, which is almost always the arxiv version. Thus it is extremely rare for me to read a journal version of a paper even if it exists.
3. Journals can provide an additional quality signal on top of which conference the paper appeared in. This seems superfluous and a lot of effort for very little signal.

I have only submitted to TCS journals when invited. I have found the process tiresome, since it creates extra work and drags on for years. Thus I am not inclined to submit otherwise. Is there a good reason for me to submit?

Answer 1

The process of journal reviewing shakes out bugs. Conference reviewers tend not to look at papers with a fine-tooth comb; the program committee process gives them too many papers to review in too short a time to do that. The main things that one should get out of a conference review are whether there are obvious flaws (these are rare) and whether the results are interesting. A journal review should evaluate those things too, of course, but it also needs to check the details of the proofs to make sure that they all work and are clearly and correctly stated.

If you are treating your journal reviewing tasks as if they were conference reviews, and only giving them a high-level interestingness check, you are doing it wrong, and depriving journal authors of a valuable service. If you are treating your conference reviews like journal reviews, spending the time to go over them carefully, and making sure that papers don't get accepted until all the details are correct, congratulations and thank you for doing such a good job, but I don't think that's the case for most STOC/FOCS/SODA program committee members.

Secondarily, many conferences still have artificially small page limits, and the journal provides space to fill in the details. The preprint version does that, too, but doesn't provide any assurance that someone has checked it carefully.

If I see a paper published five or more years ago in a conference, making claims that are not backed up by full and clearly-written detail in the proceedings, and with no journal version, I am likely to consider it untrustworthy. Is that the future you want for your research?