Talk:ModR/M

Edit conflict

@Drmies: Regarding your recent revert of some of my edits (which left me in an edit conflict state), you say "IP editor forgot to include the independent secondary sources". Er... what are you referring to?

I agree the article isn't wonderfully sourced, but the sources for the changes (especially the deletion of the laughably wrong statement "A REX prefix can optionally allow the SIB byte to use SSE registers", which appears to be someone very confused about extended registers R8-R15) are the same sources as are already in the article: the Intel and AMD processor manuals. Now, these are WP:PRIMARY sources, but per the Wikipedia policy quoted there (point 3), a primary source may be used to provide "straightforward, descriptive statements of facts" on the subject. (Context omitted but does not alter the conclusion.) How the instruction encoding works is such a thing.

This is exactly the sort of thing where a primary source is appropriate: "Person X said Y". A press release from person X saying Y is the best possible source for this; all additional sources can do is mangle the quote. (Not that I can't find a zillion and a half non-primary sources for statements about something as important to the Internet as x86 instruction encoding. One I'm finding useful is https://wiki.osdev.org/X86-64_Instruction_Encoding)

I'm happy to respond to complaints about better sourcing, but given that my changes are a clear improvement and don't make the sourcing situation any worse, It seems more productive to work on a monotone progression toward a better article without the need for backward steps.

Remember, the basic WP policy on sourcing is it's required for challenged statements. Are you challenging the factual basis of anything I added? If not, we're at "nice to have" not "absolutely required". And "nice to have" can be added incrementally. 97.102.205.224 (talk) 01:36, 19 August 2024 (UTC)

• Huh. First of all, anyone can see that you left at least four paragraphs in there without a citation. As a reminder, citations FOLLOW the material that comes from those citations. This includes the final two paragraphs, "There are, however, two exceptions:". You got back to work, and it's still in the same condition--there may be a dozen paragraphs without final citation. As for primary and secondary sources--the article doesn't have any secondary sources, really. One wonders why a product description has encyclopedic value. Drmies (talk) 12:28, 19 August 2024 (UTC)

@Drmies: It seems there are two issues here. First, I added a bunch of non-cited text. That was just laziness on my part, but not error: the added statements do have sources, and the sources are those already in the article. It seemed reasonable to postpone inline citations to the final "i-dotting and t-crossing" stage after I'd beaten the prose into shape. (It's awkward to save a WP article mid-edit other than by publishing it, and there are advantages to small commits, so I try to publish whenever the article is in good-enough shape even if I'm not done. The main thing that trips me up is if I decide to edit a section and discover too late that I need corresponding changes elsewhere in the article. Then I just save what I have and try to make the necessary repairs quickly.)

Second, the encyclopedic value of a product description. That's an m:Inclusionism debate which I'm not likely to resolve today, but I can say that, due to its widespread use, the vagaries of x86 instruction encoding have numerous WP:NOTABLE downstream consequences in fields like computer security. See, for example, A new instruction overlapping technique for improved anti-disassembly and obfuscation of x86 binaries.

Precisely because the primary sources are clear and absolutely authoritative, there really isn't any secondary analysis of the encoding itself. This isn't history where there are multiple competing sources to integrate. There's plenty of secondary analysis, but it's on the consequences of the encoding: implementation complexity caused by the encoding (in terms of silicon area and FO4 gate delay), various techniques for pre-tagging instruction boundaries in caches or creating pre-decoded instruction caches (μop cache or trace cache), the aforementioned computer security implications, etc., etc.

I started editing the article in my usual way: I came to Wikipedia for the answer to a question and found it wasn't there. (I was trying to find the number of bytes required to encode an [ESP+4] operand. The answer is 3: ModR/M + SIB + disp8.) Since I had to search farther afield for the answer (in this case, I just followed the article's existing references), I added the information to the article.

It was the fact that I used the existing references (I did, you may note, update some URLs to fix link rot) that led me to not add the additional <ref/> tags immediately; it was just a matter of cut and paste with reasonable frequency but not really critical.

I don't mind being nagged about the references, but the wholesale revert seemed a little drastic. WP:V policy does not require inline citations except in some special cases (thus why I asked if you were challenging the accuracy of the material), and it was clearly verifiable, so I was quite taken aback. 97.102.205.224 (talk) 21:10, 19 August 2024 (UTC)