Re: LYNX-DEV Lynx 2.7.2 report

[David Woolley]

> 
> When dealing with tables, if a table row contains cells which the contents
> of cell is aligned with <center>..</center> pair, the table row is split
> into several rows, which shouldn't.  For example,

This is what one would expect given the rendering rules for table in lynx.

<CENTER> is an alias for the more correct (HTML 4) <DIV ALIGN=CENTER> or
the even more correct version using styles.   It is therefore a block 
element and should have a line break before and after it.

Lynx maps </TR> to <br> and inserts a space on </TD>.

Anything better requires extensive rewriting of Lynx, including two
pass rendering (or parse tree then render, which is really the same thing).
Lots have asked for this but no-one is willing to do it.  It is likely to
have a serious impact on size and will produce ugly results unless also
combined with sideways scrolling (which will cause problems for a
significant class of Lynx users, itself).

> <TABLE>
> <TR><TD WIDTH=144><CENTER><A href="m1.jpg"><IMG SRC="m1.gif" width=115
> height=75></A></CENTER>

This looks like an abuse of tables to simulate a page description language,
anyway.

> Caching
> =======
> 
> When toggling between raw HTML and rendered HTML views, it should not
> reload the page from the original site, if cache is large enough.

Lynx caches rendered pages, not raw input.  Other tools (CERN, Squid,
Apache) exist for maintaining web page caches externally.
> 
> URL support
> ===========
> 
> File URL:
> ---------
> 
> It is possible to show someone else's home directory (if permission is

Did you mean "I would like it to be possible..."?

> granted) with following syntax:
> 
> file://localhost/~user/dir
> 
> where '~user' is automatically expanded to absolute path of 'user''s
> login directory.  Similarly, if a password is required, use the following
> syntax:
> 
> file://username:address@hidden/~user
> 
> It will do a rlogin to local machine, then go to home directory of 'user'. 

If a users wants their files to be accessible in this way, they can let it
be done by ftp.  To do otherwise considerably complicates the program for
a very limited number of users.

> FTP sort
> --------
> 
> Support for sorting by multiple criteria.  Eg:  first sort by filename,
> then date, then size. 

Lynx renders the page incrementally, even if it doesn't display it
incrementally.  You can't (in any system independent way) delegate the 
sorting to the remote server.  I don't think you will find that any
of the common proxy servers do this otherwise.

So this adds considerable complexity for a low priority function.
> 
> Searching type
> --------------
> 
> Possible to search by regular expressions.

Could be done, but would have to be turned off by default, as this is one
of the things that causes most difficulty for non Unix specialists.

> Keypad Mode
> -----------
> 
> When 'Links are numbered' or 'Links and form fields are numbered' are
> used, an option to show the numbers on the status line instead of showing
> it in the page contents.

WHY?  The two reasons for numbering links are so that you can go to them
by keying in the number, and so that you can print a table of references.
In neither context does it make sense to have to place the cursor on the
field first.

> 
> Line edit style
> ---------------
> 
> Delete key should be mapped to Delete current char by default.

This sort of change will just confuse long time users.  You can easily 
change it on a per site or user basis.

> Able to change key bindings without recompiling.

I thought you could.

> 
> If lynx is in line edit mode, press Ctrl-c should only exit line edit
> mode, not exit lynx.

I must admit I find this policy irritating, especially on Unix, where ^\
(standard key mapping) is what you would use to kill a really errant 
program.

> 
> Press right key goes to next document instead of select link.

No-one uses LINK REL=NEXT, so I don't see how you would determine the
next document.  If you mean the MSIE behaviour, you will just confuse 
lots of existing Lynx users for no great benefit.

> ^Z won't interrupt a connection/transfer process.

Is this a request to disable ^Z in the only place where it is useful???
Explain (most of your wishes are statements of behaviour after the changes).