On 3/15/21 9:02 PM, Jacob Bachmeyer wrote: > that is not currently technically feasible because WebAssembly does > not offer the APIs that most GNU packages use
Of course not, and it doesn't matter; it wouldn't make sense to expect it to use those APIs even if they were available. That would entail reliance on the local machine's resident system to perform essential services e.g. to manage the user's files and the files used by the system itself. It is the responsibility of the system living in the system image, as with all other systems, to implement the POSIX API--or any other API it wants to offer to programs that will be run within the image. (Given that we are discussion GNU, however, then yes, that means POSIX.) Expecting otherwise is a critical failure in understanding fundamental architectural issues. > One of the rationales presented to me (off-list) for this was that a > WebAssembly port of GNU could be run as a web app and therefore be > "always up-to-date" Despite quoting the salient parts from The JavaScript Trap, you have regressed to committing the same error of critiquing the computing model of traditional web apps, which is, once again, totally irrelevant. It is neither here nor there. Here you do it again: > Web apps stored on "the cloud" are bad [...] Porting to "the Web" is > simply not practical or appropriate Please, please stop using this kind of sleight of hand to redirect the context to web apps and "the cloud". "The cloud" and "the Web" are _simply_not_relevant_ to the computing model described above, which treats the browser as a runtime which can be targeted during compilation and which you happen to get "for free" on upwards of 90% of personal computing devices, *NOT* as a thin client that you all keep insisting on. It's as if there's a short-circuit in at least half of respondents' brains that prevents them from engaging in any way without at some point insisting that this *MUST* involve cloud architecture and SaaS-like web apps being the central focus. It is _absurd_ that it takes this much energy to continually refute this over and over. Ideally, it shouldn't have to occur even a single time, but failing that, once should suffice. At this point, I have to wonder how many times this has to be pointed out? Is there any number which would be sufficient? Don't want to hear it from me? Then read RMS's words from The JavaScript Trap, which has been referenced here multiple times by now. Fuck it, let's go for another: self-contained [...] you can copy it to a file on your machine, modify it, and visit that file with a browser to run it -- Colby Russell
