Yes I agree with this, as long as it’s made explicitly clear to the reader that a prior publication already exists. However I disagree that a translation is CV worthy. If this were the case, I would just translate all of my papers into every language I could and pad my CV with hundreds of copies of the same publication.
The question is, would you be able to? Translating a paper is not just passing it through google translate. It requires a great deal of work, and it has tremendous benefits for the scientists that speak the target language better than the original. But if you are able to do it (as in, the paper should pass peer review in the target language), by all means, go ahead. I'm reasonably fluent in english, and I'm obviously decent in my native language, but I haven't been able to produce a coherent translation of my own works yet (in either direction). Believe me, I've tried. I admire the few people I know who can do it. If you can... I admire you too.
Any person interested in my publication list (employer, tenure committee) would see through this ruse.
That's the point, you are assuming that it is a "ruse" and that I would try to hide the fact that it is a translation. I wouldn't. I wouldn't just list "original paper" and "translated paper" and try to make them seem different. One of them, the most "important" one (whatever that means), would get its space in the CV as the new work. The other one would get a subtitle below that one ("Also published as 'target language title' in 'target language journal'"). Btw, I do this as well with conferences/papers, i.e, if I write a paper and present it at a conference, workshop (or vice-versa), I don't try to pass them off as different works. As I said, if I value translations so much, there is no way I would try to hide that I made one.
I certainly wouldn’t reject a paper that is based on a presentation that you did at an internal seminar.
Just out of curiosity, where do you draw the line? What if the seminar is open to the public? What if you presented at the "internal seminar" of another institution? What if it was a minor conference, or a conference at a minor university? Or a conference that didn't publish the proceedings? What if they did publish the proceedings, but just hidden somewhere with low visibility, or even behind a paywall? Would the quality of the research have any bearing on your decision? I'm aware that you don't need to have a line, and I'm not implying that not having a definite line means that there is no distinction (that would be a fallacy), but you do seem to have a pretty strong opinion and a very well defined line. You can bet that at any chance I have, I /will/ try to submit papers to conferences in my home country, not because I want to "pad" my CV, but because I want to do the very little I can to change the "status quo" (a lot of researchers from my country wont publish locally, because doing so would disqualify them from publishing in an "important" venue, which in turn makes the local venues even less important. The "important" venues are obviously very pleased with that arrangement). Whether to accept the "duplicate" presentation should be a decision of the host venue, based on whether the quality of the paper warrants another talk and whether the target audience is likely to have been exposed to that work and/or will gain by being exposed again. To disregard it without that consideration is, IMHO, irresponsible towards the goal of disseminating science.
As a physicist, I’m only interested in the physics aspect of your paper. As a computer scientist, I might not understand your physics notations. As a result your paper will probably just drift into obscurity in all venues. The best approach is to write three different papers (not just minor tweaks, but tailored for the audience), each with their own unique take on the problem, each with unique results (and each of course referencing the other publications).
And now the reason why I didn't reply to the rest of your post. It seems that we are mostly in agreement: the difference is that I consider what you list also to be a duplicate submission: there is only one "research", you are not "extending" the work to add new stuff for the other journals, rather, you are hiding part of the results from the other journals (with good reason). So if I work in more than one paper (I just "solved" one of the physics problems! Yay!), I will not consider them to be different works. I would even classify it as "salami", if that weren't a necessary evil given the three different audiences. So, naturally, I assumed that in your view they were the also same as well and that you would reject the second one without a second thought. I still disagree with your statement that the benefits are only "very few", and with coherence of the original claim ("the goal is to disseminate knowledge... so if I find someone publishing more than once, I will reject it")
It’s always been the case that the authors tried to hide (by omitting references to old work) that in the past year there are no new results to publish, and they’re trying to pass off an old paper as a new one.
But that's really the problem, isn't it? Is not the so-called "self plagiarising" itself. The problem is that researchers commit fraud to pad their CVs and reports because the metrics used to measure them are lacking (but oh, very convenient to the publishers). You are measured by how important are the journals you publish in, rather by the impact of the research. A much better measure would be the raw number of citations or a derivative of that measure (and if it happens that you get higher number of citations by going to a high impact journal, go there by all means). I've never opened an issue of "Nature" to see if there is something of my interest, unlike with specialised journals (I have just followed links to it)... yet a paper in Nature would be worth more for my CV that one in those that I actually consult and cite from. The problem of authors trying to pass old works as new (and several others that are actually more damaging) would become irrelevant if the metrics didn't assign a high value to the "number" of publications. In absence of that metric, publishing a paper in more than one journal, even if it is verbatim, would lead to one of two outcomes: either the number of citations increases (in which case, it increased awareness of the paper, thus serving to disseminate science), or it doesn't (in which case it should have zero effect in your CV, regardless of how many times you've tried to publish it). You seem to respect the metric, and actions that put it at risk are unethical in your view (that's why I accused of siding with the publishers. I apologise for insulting you). I despise the metric and believe that it is damaging, so I just play the game when I have to. Then again, I'm just a phd student with very few results (some of them with tiny "also in..." subtitles), I've never been in a position of hiring someone else, and my only instances of evaluating someone else's performance have not been based on the length of their CV but in the papers I've had to read.
(I didn't really have any coherent place where to put this, so I'll just drop it here. I'll grant you the point that the reviewers' time is a finite resource. You are, obviously, correct. But I don't think that with sane editors there would too be much waste.)