References going to unexpected entries

[ecm-pushbx]

During my work today on #6 I found that AH=4Bh and AX=3306h references sometimes(?) led to massively less popular entries (unpopular 4Bh vs EXEC, unpopular 3306h vs true version).

The 4Bh case is fixed by making the unique ID 214B--DX0000; I didn't find a template for how to indicate DS:DX=all-zeroes for that one.

The 3306h case apparently can be worked around by referring to AX=3306h"DOS" so I dutifully changed references this way. (Maybe there's some I missed, from other INTs perhaps?)

Interestingly EXEC is "(AH=4Bh/CX=04h)" and the unpopular 4Bh is "(AH=4Bh/DS:DX=0000h)" on https://fd.lod.bz/rbil/zint/index_21.html - The EXEC CX= is bogus, but it's interesting how the unpopular 4Bh has DS:DX=0000h which I don't see reflected in the source files I have been working on.

[ecm-pushbx]

https://fd.lod.bz/rbil/interrup/other_os/2120.html links to

The PTS OEM code does not correlate with the DOS OEM code returned by INT 21/AH=30h, likewise the PTS OEM revision does not correlate with the DOS revision returned by INT 21/AX=3306h.
SeeAlso: INT 21/AH=30h,INT 21/AX=3306h

The first (notes) AH=30h incorrectly links to the index, the first (notes) AX=3306h is fine, the second (see also) AH=30h links to another (new) unpopular page, and the second AX=3306h also links to the unpopular page.

[shidel]

Yeah, I know. I have seen that sort of thing as well.

The old RBILtoHTML parser did a miraculous job creating all of those cross references. But, because of how most of those
were inserted into the text, it was probably about 70% or more of the time it was just guessing on where to go. That whole cross-linking logic it used was an experiment. It was actually kind of amazing it every worked anywhere near as well as it did. But, if you followed enough links, you would absolutely eventually end up somewhere unexpected.

There are a number of ways to improve the cross-linking. Plus, there is also a bunch of things that are not going to be needed anymore. Such as on-the-fly corrections and additions to the text.

While the old one can stand for a little bit, I plan on creating a new one from scratch. It will have a lot of improvements, including it it's cross-linking reliability. Also, I intend it to throw warnings for things it is unsure about. Because now, we can simply update those links to remove any ambiguity.

As I mentioned in another thread, it is time to retire that old parser an favor of a newer one. When I do, that old project will be archived and the new parser will live in this project. So, it does not hurt to keep track of the old versions failings here. It will us good examples to test against.

:-)