Re: LYNX-DEV libwww5

[Klaus Weide]

On Mon, 21 Oct 1996, Foteos Macrides wrote:

[to Walter Skorski:]
>       You need to do a global search for HadVMSInterrupt and reproduce
> my for-VMS code for Unix.  I just did a search and counted 14 LYfoo.c
> modules and one libwww-FM module.  You can add the prompt in LYClean.c,
> but need to handle the reply in indifferent ways in various modules,
> and deal with the input buffer changes associated with the procedure.
> How you go about it, and whether the existing code for VMS would need
> changing, depends on whether a libwww upgrade, and serious use of its
                       ^^^^^^^^^^^^^^^^^^^^^^^^
> select()-based threading, are on the agenda for Lynx. 
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^
Fote has mentioned this several times recently, so I assume that 
he would like to see a discussion on this.  Ok, so would I.

Some pros and cons:

PRO

- Once upgraded to new libwww, can reuse existing code for new 
  protocols (like HTTP 1.1 & PICS & PEP...). 

- Easier to plug in additional modules that come with libwww (which may
  not be wanted by all, but useful for some people; cache, hostname
  cache,..) No need to reinvent wheels all the time.
  (New wheels still have to be fitted in).

- As Fote mentioned, select() based multithreading (would allow Lynx
  to e.g. scroll through document while loading another one in a 
  cleaner way than the recently posted hack, Lynx can do several
  retrievals at once..)  No need to make use of this though, could be
  added later.

- Might help conpiling for Win32 or Mac.  (my idea..)
  New library has support for Windowing system's event loop, include
  files etc.

- May help going to (GNU) configure script for installation.

- It was said once that it would make real table support easier.

CONTRA

- May increase size significantly.  ??? currently unknown. 

- May be slower ??? (probably insignificant, maybe just the opposite.)

- Some drastic code changes needed to do seriously.

- Needs work.   Someone is hacking around with it currently (me), but
  that doesn't mean it will lead to a useful version.

- Maybe too little incentive.  Not enough (immediate?) benefits to 
  implement and get tested.

THOUGHTS?

  Klaus


;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;