Re: lynx-dev THANKS AND QUESTION

[Klaus Weide]

On Mon, 3 Jan 2000 address@hidden wrote:
> Oops.  Tripped up in an inconsistency.  I usually argue the minimalist
> POV; I believe when source caching was debated, I took the side of
> leaving caching to a proxy.  But, yes, when it became available I
> turned it on -- for better or for worse, the development effort had
> already been spent.

So have I.  And after all I fixed a bunch of problems with it recently.
Until other problems with it (old and new one(s)) made me turn it off
again by default.

> You're provoking me to reconsider.

Turn it on or off as you like; just don't assume that whatever you do
is The Standard Way.

> But, as long as you acknowledge bugs in the facility, I deem it one
> of the bugs that d(ownload) bypasses the source cache.

You are abusing the word "bug" here.  Yeah it's fashionable to call
everything from missing features to features-behaving-differently-from-
what-I-like "bugs".

There are enough real bugs to discover, no reason to inflate the number
with opinion bugs.

Leonid and I recently talked about this, i.e. whether 'd'ownload should
use a source_cache copy.  I said then that yes that would make some sense.
But I've changed my opinion again now...

'D'ownload means "download", period.  It's always meant that.
Consider the meaning of the word.  The user can expect that a new
download attempt will be made, rather than just using a (possibly
corrupt) cached copy.

If you want to argue for a new 'D'ownload that doesn't always mean
"download again" you should call it something else.  Maybe
SOFT_DOWNLOAD.  Different function, different name or key for KEYMAP.

   ---

Your formulation that "... d(ownload) bypasses the source cache",
specifically the word "bypass", seems to show that you have a mental
picture where taking from the cache is the normal thing, and it takes
some special action to "bypass" this.  That's far from how it really
is.  You don't need to "do" something in order to bypass the cache.
You'd have to implement something special in order to use the cache.
(IOW it doesn't really act much like a cache should, I haven't kept my
opinion that this is fundamentally flawed a secret.)

   Klaus