lynx-dev Re: file-upload v0.01, also re: suggestion for --enable-color-style

[Klaus Weide]

On Fri, 4 Jun 1999, Rob Partington wrote:
> In message <address@hidden>, 
>            Klaus Weide writes:
> > 
> > I know why.  Since Tom and I took your original color-style code and
> > merged it with the then current devel code (at around 2.7, I think), and 
> > then tried to make it work - there were numerous inconsistencies and 
> > crashes, with nearly no useful comments, no instructions or documentation
> > from you - you didn't participate at all in any of that.  In fact, I don't
> 
> I don't remember getting any emails about the internals of it.  Perhaps
> if I had, I might have been able to help.  I've pointed out several times
> that I just don't have enough time to follow every message in lynx-dev;
> if I missed messages asking for help with colour-styles, I apologise.

At that time - a couple years ago - you didn't seem to follow lynx-dev
at all.  There was an extended period of time, before we integrated the
"alpha" style code, when people tried to get you code running and
questions remained unanswered.

It just didn't seem likely you would honour a request like "please add
comments to all added code parts explaining how they are supposed to
work together"; maybe I should have tried.  :)

> On the other hand, if you didn't contact me, it's hardly surprising 
> I didn't participate, is it?  

It *was* surprising, given that you had just offered the code and
indicated that you were working on improving it.  At least I don't
remember any announcement that you wouldn't participate any more; I do
remember at least one message to you directly that never got a reply.

> > remember ever seeing *any* indication that you have even looked at the
> > merged code since then, as far as color styles are concerned.  You just
> > discouraged the effort, and said you had a "completely rewritten" and
> > "completely different" version:
> > <URL:http://www.flora.org/lynx-dev/html/month0597/msg00684.html>.
> 
> Because that colour-styles code is a deadend.  It was a proof of concept
> hack to see if it were possible, nothing more.  I'm happy that people
> considered it good enough to put in the mainline code, but I have _never_
> thought it was good enough or useful enough to inflict on other people.
> 
> > But that new-and-improved version never appeared, I am quite sure you
> > never announced something like it to lynx-dev since I was waiting for it,
> > and the "pre-alpha code" remained all we had to build on.
> 
> Unfortunately, it mutated into ObjectTreeLynx.  Again, a proof of concept
> hack that may or may not be useful to other people.  Still a long way off
> though because inconvenient things like jobs, moving house, etc. get in 
> the way.  My apologies.

Well, you seem to confirm that you never released colour-styles (or
something derived for it) in anything beyond "proof of concept hack"
form.  So whether it was a deadend or not, if we hadn't taken the
pre-alpha code, there still wouldn't be colour style support at all in
the mainstream lynx (unless someone would have started it from scratch).

> > I have never seen a version that worked (or compiled) with color styles
> > and slang.  AFAIK, nobody except you has ever got this to work, with any
> > of the code you published or any of its descendants.
> 
> I published the code I had working.  Same as the form-based file upload
> I've just uploaded the patch for.  It works for me.  If people want to
> integrate it into the mainstream Lynx, fine.  I'll help with debugging 
> if I'm asked.
> 
> > It took a lot of reverse engineering (over a long period of time) to more
> > or less understand how the color-style code (in that documentation-free
> > "pre-alpha" version, since that's all we had) was supposed to work with
> > (n)curses, at least for me, and there is still stuff I don't understand. 
> 
> See above.  I don't remember any emails asking for information.  
> 
> > > I know it works, I wrote it and I use it.
> > 
> > You must be talking about something you never shared with lynx-dev.
> 
> I shared the code I had working.  I apologise if my code didn't work for
> you, but it did work for me.  I'll see if I can dig it out when I get home.

If that's still from around 2.6 or 2.7 times, it would take someone *very*
dedicated to try to apply the necessary changes to the current code.

It's all water under the bridge.  I can't complain that you dropped out
of lynx-dev without (IMHO) adequate warning - I do that myself.  It bugged
me that you had promised a better version which then never appeared - but
it was our decision to take the pre-alpha stuff and try to do something
with it.  I'd be wary of repeating the experience though, now with the
form-based file upload code.  Your diff is against 1.8.1dev.16 (you
wrote "iirc", so not even that's for sure).

[from your announcement:]
> [It doesn't seem to work against 2-8-1 for some as yet unknown
> reason, I'll look into that later today]
>
> Basically, it's alpha code, and it works for me in some
> situations.  I'll fix whatever bugs people point out as best
> I can.

I think it's very unlikely that anyone will point out any bugs -
who is going to still have 1.8.1dev.16 around to apply your patch?
The current version is 2.8.2rel.1.  There are lots of changes in
at least one area where your patch also makes lots of changes
(HText_SubmitForm).  If you cannot work from the current code and
supply a patch against it, I'm afraid we'd have a similar deadend
situation all over again.  If somebody else took enough interest to
make your form-based file upload code work with 2.8.2, he or she'd
end up with code you are not familiar with, and I doubt that your
offer to help debug and fix bugs would help much - even assuming
you'd have time for that though you don't have time to update to
a current version.

Well, I looked through the code, not in detail but enough for some
observations:

You use various macros -DRJP -DBASE64 -DRJPBASE64 and OLDLYNX, the
intended meaning is nowhere explained.  This reminds me very much
of the colour-style code...  Could you please provide some
documentation for this?  Otherwise we'd definitely be back to a
reverse-engineering situation.

There seem to be various historical layers of code, not all of them
in working order, in some way selectable by the macros.  Again this
reminds me of colour-style.  Which code is dead and which isn't?

Could you please provide some comments how it's all supposed to work
together and what it does.

You add new functions HTSABcopy, HTSABCat.  Maybe we already have
a near-equivalent of at least one of them (MemAllocCopy) that could
be reused.

Trace output is written to stderr, unconditionally.
(It makes me wonder, are *you* really use this code?)

There seems to be an arbitrary file size limit (65536 bytes?),
effective for at least some combination of the preprocessor macros,
and no warning or error when this is exceeded.  Also 65536 unnecessary
bytes are alloced for all form submissions, without any need in most
cases.

The code is using unix I/O functions for file access instead of
stdio, and without proper error checking.

I think the user should be able to change the filename, but that
doesn't seem to be the case.

Large files should be handled more efficiently, without passing and
copying the file contents around as <some struct>.post_data of which
there can be several copies.  (Yes this would require significant
changes to the way POST is handled throughout.)

Could you please provide some blurb how the user interface for
file input fields works.  I don't mean necessarily finished Users
Guide additions, but there should be *something* *in* the patch.
You may have written something about it in the past to lynx-dev,
I don't know.  You have a short text on <http://lynx.browser.org/rp/>,
but...

The page at <http://lynx.browser.org/rp/> is definitely out of date.
It only has a link to a year-old previous version, without any indication
that a newer one is there.  It has text like "instructions on how to do
this later" and "This patch is currently broken" and "I 'll try and get a
fix up sometime today or tomorrow" - THAT WAS APPARENTLY WRITTEN ABOUT
A YEAR AGO!

Sorry if this appears too critical.  But I think you should know.
Don't say nobody asked.

   Klaus