Subject: Re: Commit: some GnomeVFS support
From: Mike Nordell (tamlin@algonet.se)
Date: Sun Jan 21 2001 - 12:22:55 CST
Martin Sevior wrote:
> On Sat, 20 Jan 2001, Dom Lachowicz wrote:
>
> > AbiWord now (optionally) has GnomeVFS support in some of its exporters
> > (namely, all of the text and XML exporters). Compile with
ABI_OPT_GNOMEVFS=1
[...]
> This is a really cool feature!
However cool it's IMNSHO wrong.
If we want to do this, we should create an interface/class-hierarcy all
platforms could use/extend/implement. To do this just against one platform
and one library is IMSHO plain stupid.
Ask yourself, do we *really* want to introduce yet more dependencies on
stuff? I'd say no. We need to create our own interfaces against this kind of
stuff to leta any platform implement it in any ways suitable or we shouldn't
do it at all.
What we instead should do is create some abstract class (hierarchy) to
read/write streams of data for our im-/ex-porters and from that point we
should then implement the other "fancy" stuff, like HTTP D/L. Please compare
it with C++ iostreams. The data provider is the strmbuf.
This does not even take into account the bonehead comment of Jeff Hostetler,
27 January 1999 (according to our coding guildelines document #16). The C++
library was put together by many *experts* on this area, give these
*experts* the benefit of a dubt before rejecting their hard work stating "I
know better!".
I just wanted to say I think - no, I'm *convinced* - that this is the wrong
way to do it and it only promotes a bloated monolithic unmaintainable mess.
Lets implement a stream to read/write from and then let platforms implement
it in whatever way they want to (mmap(), fread(), read(), I don't care, but
I *do* care about being locked in by this design).
/Mike - please don't cc
This archive was generated by hypermail 2b25 : Sun Jan 21 2001 - 12:23:23 CST