[chmspec-devel] $FIftiMain

[Razvan Cojocaru]

Hello.

Pabs, here are the things I think the $FIftiMain section of chmspec 
could use or that I thought lack clarity.

As a minor observation I think that the way to decode integers such as 
the big endian ENCINT and the scale/root integers should be presented 
closer to the places where they appear in the file structures, not the 
1st thing in the file.

That being said, we turn to Table 5.19.
I could not tell that I should start with the node pointed to by 
location 0x14 (DWORD). "Offset to the last word tree block" doesn't tell 
me it's the one I should try first. Also, the description of what the 
depth of the tree blocks are seems insufficient to me, and I had to look 
at the chmdeco code to figure out what you meant. My critique to this 
kind of things is basically that you should be more detailed. They 
probably seem obvious to you since you reversed engineered it but it's 
not that obvious to people who haven't had such low level experience 
with the chm format.

Then you keep talking about what follows in the file (WLC entries) and 
describe them. That's all well and good but at this point you haven't 
introduced index and leaf nodes and the concept of a WLC means nothing 
to me.
As a developer, I was interested in the steps to take to make a search 
happen using $FIftiMain but from the spec I could only figure out some 
of the layout of the $FIftiMain file. I think the section could use a 
reorganization about the way the code would flow to get the pages of a 
word: read the index node, iterate through the entries, get the leaf 
node, go to WLC, etc.
Also, an example of how words are represented would be useful. More and 
clearer examples would be welcome. Not complex stuff, but like we talked 
on the xchm-general mailing list - like word1 is animal and word2 is 
antitrust, and the word in the second leaf is titrust, starting at 
position 2 in the previous word, etc.

The windows.chm example you presented doesn't quite illustrate many of 
your points. I think a fictitious example would have perhaps illustrated 
the concepts better.

Hope it helps. I'll think about it more when I get some time and see if 
I've left anything out here.

Cheers,
Razvan