Message-Id: <36FBE6A6.D1407A64@cableol.co.uk> Date: Fri, 26 Mar 1999 19:57:26 +0000 From: Allens X-Mailer: Mozilla 4.04 [en] (Win95; I) Mime-Version: 1.0 To: djgpp AT delorie DOT com Subject: Re: DJGPP: the future is... ? References: <199903260517 DOT AAA32193 AT envy DOT delorie DOT com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Reply-To: djgpp AT delorie DOT com Wow, thats interesting. *my* comments and opinions below. (Sorry about *big* mail) Peter Allen DJ Delorie wrote: > > Don't panic! > > I've been doing some thinking lately about where DJGPP is going. I've > concluded to myself that it really isn't *going* anywhere, it's > *there*. The core code is pretty stable and feature-full. GNU ports > come out often enough, and we're becoming "just another unix platform" > for a lot of people. The web pages have pretty much everything I can > think to add to them (although we can always use more documentation :) > The big focus these days is on third-party additions (like Allegro) > and applications. > > Cool. > > The big question I have for you all is this: What's next? > > I don't mean "Let's write application XYZ", I mean for the DJGPP > project itself. Think BIG. I don't see too many people installing > DOS these days, and Cygwin is shaping up to be almost as good as DJGPP > (hey, I can boast - I'm on both teams) so the Win32 systems will see > less demand for DJGPP over time (I expect, at least). Plus, a lot of The os that is used as the primary os is going to be a big descision. I would definatly vote for sticking with dos, as windows support would be a huge rewrite, and IMHO I don't want to have to do windows anyway. ( I know that windows console is very similar, but its just not quite right ) I think we might think about expanding the support for other languages and api's more, although a lot of the things have already been done. (i.e. mesa3d devel, the fortran part of the compiler, and the pascal, etc). On this theme I think we need to continue supporting as many different platforms as possible for things like allegro, and other common addons. > DJGPP users are switching to Linux or WinNT/Cygwin. Because of cygwin for win32, IMHO I don't think we will get many new users there. The main reason for windows support is so that existing users don't have to download cygwin to do win32, and have the added complication of having two gcc's etc on one system. > So what are our choices? > > Well, we can try to saturate the DOS market. I don't know how to do > that, unless we spend a fortune on ads in PC magazines :-( Perhaps an > effort to "spread the word" in other forums (nicely, please) would > "enlighten the uninformed", but it's for diminishing returns. Agreed. Just out of interest, where does something like cygwin do its advertising? As they are trying to attract a reasnable similar market to DJGPP, so we could try advertising in the dos equivalents. Obviously DJGPP has not got the commercial backing of cygwin, so we cannot expect the same publicity budget. If we were going to advertise, I reckon we need to go for very specific audiences, like finding somewhere to addvertise for people dabbling in C for the first time, as this is where DJGPP has an advantage in being free. Hopefully once started with DJGPP, new users will stick to it. The other area where we can advertise possible is people porting unix console apps to dos, as DJGPP is (should be anyway) nearly 100% unix gcc compatible. We might have problems with doing this and DJ's licence agreement for his job though. > We could try to make DJGPP a Windows-native system. I don't think > this is a good idea because 90% of the value in DJGPP is the way it > hides DOS, and we'd have to throw it all away and start from scratch > if we switched to the Win32 API. Plus, Cygwin already does Win32, and > legally I can't promote such a project because that's what I already > do for Cygnus. If you like this option, join the cygwin team - you'll > be much happier, and we can always use more help. Agreed. IMHO DJGPP is not and never should be biased towards windows. Yes have windows extentions, because they can do exactly the same job as a native windows compiler, but without being part of windows. Also, we have enough trouble with windows bugs in dos, so think what it would be like if DJGPP was native windows :) > For the same reason, DJGPP for Linux is a bad idea. Heck, DJGPP is a > port of the Linux tools themselves! Agreed > > We could overhaul DJGPP again for ELF support and a few other > fundamental design changes, but why mess with a good thing? Sure, > we've got a list of bugs to fix (like C++ templates in COFF) but > they'll get fixed eventually. Such redesigns would have little real > effect on the project. True, and COFF does have a lot of similarities to PE, which is just glorified COFF, so that enables an easy (ish) extention to PE. On that note, COFF is very well established, so it is probably easier to convert to new binary formats than ELF, which although is great for UN*X it's not nearly so good for other platforms. > > So, I'm at a loss as to where we should be focusing our energy at this > time. Mailing list traffic doubled every year from 1993 to 1997, but > *dropped* 10% in 1998 (1999 isn't looking too good, either), and I have noticed a considerable drop in the number of installation problems recently. That is good. The installation *could* be worked on if people are looking for things to do. The zip-picker theme is good, but what about going one step further, (along any roads on that kind of theme). Another thing I have thought we could try doing is reducing the amount of stuff that has to be downloaded. I think the modularization of different zips is good, as it allows updated without downloading the whole thing, although that could be done with patches. Obviously changing the compression format would help, but a 'different' way of reducing the size I was thinking of was to use dynamic linking of librarys for the base gcc system. (So the bits that are in no-way standalone anyway.) This does bring about some rather fundamental problems like library versions, but those wouldn't be as critical as they are on OS's where every program depends on one library. (Linux). > delorie.com's web server has had a pretty steady load for the last two > years, even though I've got plenty of spare resources. I think our > period of growth is over unless we start something new, but what? Possibly trying to establish some active sites where people are doing 'things' like group projects or something. > So my task for you, my loyal fans, devoted followers, silent lurkers, > and the occasional tax collector (hiss!) is to help shape the future! > Let's get those ideas flowing and figure out what the Next Big Thing > for us will be! > > Considerations (but not limitations): > > * GPL. We've come a long way with it, no reason to change now. Agreed. The licence agreement is brilliant for DJGPP is good because it allows commercial development, but still has the basis of the GPL. (I could be wrong on this as I haven't read the licence properly). > * Should give something to the community. DJGPP was built by the > community, the benefits should go to the community. > > * Traffic to my web server means money for us, which I turn into more > servers, bigger disks, faster net connections, time to write CGIs, > etc. The main DJGPP server is a P166 with 27Gb of disk and a 1Mb/s > link, but if you folks want to chip in for a PII/450 I'll get one > ;-) Current traffic just about covers ISP, electric bills, and > upkeep (about $9,000/year). Active projects? > * Should be long term and highly visible, so we'll all get rich and > famous (we hope) through contracts and such. > > * Should be something that can grow on its own. For example, if > personal labor is involved (like consulting), you're limited by how > much you can do, but a web page, computer program, or software > package can grow and multiply without needing more people-hours. > This also frees us up to do other projects when we're done :-) > > So crank up those brains, think carefully, and let's hear your > suggesions. No flames please! You may send me private mail if you > don't want your ideas public, else send them to the djgpp forum. > > DJ