Re: bad relative urls in texinfo-4.0f

[Per Bothner]

Eli Zaretskii wrote:
> > The extra anchor references "#XML%20tools" are redundant and ugly.
>
> Can you tell why ``ugly''?  Those links are never displayed to the user, 

They *are* displayed to the user, at least in the URL bar of Mozilla.
Otherwise I wouldn't care.

> And they are not redundant, even when there's only one node per file: the 
> file does not begin with the node text, so the direct link to the anchor 
> makes sure you land on the text, not on the preamble.

This seems like questionable UI design.  The link should land us at the
top of the page, else the user may scroll up to see what is above it.

> One problem this 
> fixes are the references to the Top node (which is on the same file as 
> the TOC):

Well, in my kawa.texi, the toc follows the contents of the Top node,
and I see no reason you'd want #Top.  Perhaps this is different for
other texinfo documents?

> > I can see that they are needed when there are multiple nodes on
> > a page but I see no reason for them when referring to the
> > main or only node on a page.
>
> The problem is that makeinfo doesn't know whether there will be one or 
> mode nodes on that file.  When the HREF anchor is generated, it could 
> well be that the node pointed to was not yet processed at all (a forward 
> reference); given the possibility of file-name clashes between several 
> nodes, makeinfo has no idea how many nodes the file XML-tools.html will 
> eventually harbor.
>
> Even for nodes which were already processed, the information about 
> whether they are alone on their file does not exist, and is not very easy 
> to compute.

Actually, I don't think we need to care whether a node is alone in a 
file, only whether it is the "top-most" node in a file.  This we could
know in advance.

> We could probably make another pass after everything is 
> finished, and remove those anchors from links to files that have only one 
> node, but is the problem really worth the slow-down and complexity of 
> another pass?

Who cares about slow-down?  I can see implementation complexity being an
issue - but slow-down shuld not be an concern.

> Is the ugliness of the %20 thing the only problem, or are there others?

Note the %20 thing - the entire #Node thing - which *is* user visible.
--
        --Per Bothner
address@hidden   http://www.bothner.com/per/