Yancey Yeargan wrote:
> 
> I'm actually very tired of every program on the planet coming
> with it's own spelling checker.  When is someone going to create
> a no-cost, cross-platform spellcheck system that we can use for
> all our applications?

There's ispell. It has only three drawbacks: (1) no robust API, and none 
whatsoever on Win32 or MacOS, (2) its dictionaries are riddled with 
errors, omissions and crap, largely to maintain backward compatibility 
with apps that rely on the bad dictionary, and (3) it is installed on 
fewer than 1% of desktop computers.

Once the first two problems are solved and it gets nice, standard and 
stable COM and Carbon APIs both for spellcheck and for augmenting 
personal dictionaries, and it gets  a trustworthy set of dictionaries, 
you've got a bigger problem. Who's going to get it onto those 94+ 
percent of computers that don't have ispell and whose OS vendors 
couldn't care less about it? It is on all new Macs, probably used solely 
for seeding Darwin's random number generator, but Windows will be a 
tougher nut to crack.

This isn't something the Mozilla project can tackle, since Mozilla is 
used by eight people worldwide. AOL/Netscape isn't going to bother, 
since they already license the same nice Lermout & Hauspie spelling 
subsystem most vendors use.

> Unix has ispell which works for my needs.  I don't need or want
> MS Word's feature of underlining a word as soon as I've
> misspelled it.  I'm just fine with spellchecking my docs just
> before I publish or send them.  I do want a single spellchecker
> instead of five or six on my computer.

Then you probably have contempt for graphical mail clients like that in 
Netscape, Mozilla, Groupwise, Notes, MS Outlook/OE, Eudora, et al 
anyway. Go on sending all your mail with Emacs and don't forget to 
dryclean that hair shirt once in a while. Realtime spellcheck is not a 
feature anyone is going to back you on removing from software. Ever. You 
can shut it off yourself, as can anyone else bothered by it. But telling 
people to stop asking for spellcheck and instead save their composed 
message to a file, running ispell on it, and then gettung the fixed 
document back into the mail app isn't going to be seen by anyone as a 
step forward.

The issue here is that spellcheck seems to have been deactivated in 
recent nightly builds, just as most of the MIME attachment conig stuff 
has, temporarily making Mozilla Mail/News utterly useless. It's hard to 
eat your dogfood when it forces you to use a different app to send 
binary attachments and you can't run spellcheck.

I'm sure this will all come back to the trunk in the next few months, 
with some fabulous rewrites that haven't landed yet, but for now, 
Mozilla mail is useless.

> We've come up with Open JVM Integration (OJI) so that we can use 
> anyone's Java engine.  How about someone creating OSCI, Open 
> Spell-Checker Integration?

Speaking of OJI, that's hardly a gem. As distributed nowadays, it 
installs its own JVM and on Unixes still has fatal path problems out of 
the box. With Java 1.3 and even 1.2, there's no reason for this. 
StarOffice, Java Web Start, the Java Plugin and any number of modern 
IDEs and app servers happily detect  available JVMs (whether through 
environment variables, a registry scan, or a filesystem search) and let 
you choose the JVM(s) you want to use, no 56343rd copy of the JRE required.

It's startling, really. Even after all these years of staff turnover, 
the Netscape client team still seems to be comprised almost entirely of 
academic Unix programmers with little awareness of or interest in using 
any of the modern APIs available to them on the host OSes. Java 
integration in particular seems to have been done in a world that hasn't 
gotten past JDK 1.1. And cross-platform code is nice, but it's appalling 
that even the official Netscape 6.x releases can't install globally 
through an administrator account on WinNT and Win2K. Do they have anyone 
there who ever developed a Windows app professionally, or are they just 
handing a copy of VC++ to a Unix person?

--
Steve Koppelman


Reply via email to