Slashdot Mirror
Neither Stable Nor Unstable: A Midrange Debian?
truefluke writes: "This was forwarded to me, from a friend who is a very loyal Debian supporter, with the cool news that Debian could become 'more up to date more often'. This news appeared here on the
Debian news source. A good idea, and prob incentive for more folks to try out Debian without resorting to the old saw of 'too old / too slow.'" Debian developer Anthony Towns says in the list posting mentioned there that "[t]his is a (mostly finished) project that will allow us to test out distribution by making it "sludgey" rather than frozen[.]" Sounds like the same logic behind the Caldera "Technology Preview" and Red Hat's Rawhide -- give people more of what they want of The Bleeding Edge, without getting reckless with the official release.
If you're after the 'More' unstable code, just find the right apt sources. There are a lot of third party apt sources that package nightly builds of various projects. ljlane's nightly CVS builds of Enlightenment-related packages, rkrusty's (nightly?) CVS builds of KDE-related packages, etc. It's out there. :)
--
"Convictions are more dangerous enemies of truth than lies."
http://www.debian.org has a link to "Getting Debian on CD" and some of the listed vendors indicate that they ship development snapshots.
In particular, I've purchased CDs from Greenbush, http://www.greenbush.com
Anthony Towns:
Supporting this, there's some Apt changes in CVS that'll let people choose a few packages from one distribution and leave the rest from another. Two possibilities come to mind: either running "testing" most of the time, but using a bunch of programs from "unstable" because you're interested in their development; or running mostly from "unstable" except for a few packages you can't afford to have break on that machine. Either way you have a slightly larger buffer between an upload and it making it into "testing".
I've been waiting for something like this to be added, I couldn't believe I was the only one who would want it. Anyway, this will almost certainly bring more testers, since you can run the fairly-safe testing distribution and then test the bleeding-edge stuff you are interested in. Today, it is basically all or nothing. You either run the unstable branch and risk it going to hell completely or you stay away from testing the packages that you are most interested in. It also might make it easier to track down bugs if the parts of the system you are not particularly focusing on are likely to be fairly stable.
Logi - I can do anything, but not everything.
Give the people what they want, and they'll come back for more. Besides, we all know that if we use something that's beta/bleeding edge, we know we are risking life, limb and sanity.
We are forwarned already, now arm us.
-----------------------------
1,2,3,4 Moderation has to Go!
Anybody whose running X 4.0.1 or Glibc 2.2 on Debian, of which I understand there is a decent number. I'm impatient enough to some times pull programs from Incoming, which means I'll be there using unstable and making sure that at least the brown paper bag bugs get reported.
What I'm saying is, what if you have a slow connection at home, but have access to a fast connection somewhere else (for instance at work). In those cases, an ISO is nice (and better than the pseudo-image thingie they have now. Tried it, didn't work).
Je ne parle pas francais.
And if you have a slow connection?
Je ne parle pas francais.
I was wondering when someone was going to bring that up. I guess I shouldn't be suprised it was you :-)
Usually what I see with debian is the opposite. Someone packages foo and CLAIMS that it "requires" libBar 1.23 or greater when in fact any libBar greater than 1.0 works just fine. But the maintainer of foo has libBar 1.23 installed and was too lazy to do any real dependency analysis. So users are forced to upgrade half the packages on their system to satisfy the alleged dependencies of a single piece of software they want to install.
Sieg Ackermann!!!
huh?
I personaly would like them to use 'nonthreatening' and 'threatening' instead of 'free' and 'nonfree'. It would simply be a lot more understandable.
:)
How so.....then you have to explain why organizations and/or companies that release closed source software that they spent time and money creating with a license of their choosing is more disturbing than labeling it as a threat because you prefer open-source software.
Being an zealot, an elitist or a xenophobe is a waste of time. Other than that, your post was on target.
Is this sort of release really neccessary?
Yes. There were several times when apt would kick up its legs and thrash in dependancy errors on me, and it would take a few days before I could fix it. A couple of times a dist-upgrade would hose perl and apt so bad I had to reinstall.
Not that I'm not a card carring Debian Nazi, even with those two reinstalls it was soooooooo much nicer than running Red Hat.
Yup. With the latest release of Debian, and hearing so much about the godlike powers of the packaging tools, I had decided that my installation of Linux on my main home machine would be Debian. The ideologican purity is a good thing (to me) as well.
But I like to play on the cutting edge - a "QA" version of Debian would be perfect.
Torrey Hoffman (Azog)
Torrey Hoffman (Azog)
"HTML needs a rant tag" - Alan Cox
So if "testing" is better (more usable) than "unstable", it begs the question, "who's going to use unstable"? And if no one uses unstable, unstable will remain just that -- unstable.
Never meant half of the things I said to you. So you know, there's a half that might be true - G. Phillips
That's what i said,
2 17&cid=38
http://slashdot.org/comments.pl?sid=00/08/23/2045
but of course, the zealouts that can't take the truth always attack... heh
You need to set the system variables ftp_proxy and http_proxy before using wget through a proxy.
I use (in NT):
set http_proxy=a.b.c.d:80
wget http://foo.bar.baz/wibble.nerf
HTH,
dave
Wow, it's a miracle that you use dselect but still _like_ the package system.
I find it much simpler to use apt, where a command like:
apt-get install perl
Will fetch and install 'perl', as well as all the packages that it depends on (excluding the ones you have installed allready). There is a console based apt that functions much like dselect as well.
Given the rock-solidness of the Debian stable releases, and the fact that a lot of people do use the unstable versions as well, this sounds as a great idea. Hopefully this won't generate a lot of overhead for the different maintainers.
You mean like this? It was great fun, but I think Freesco was better... it all fits on a floppy.
SYS 64738 NO CARRIER
Go into Borders bookstore on Tottenham Court Road and you can get all of those and Slackware 7, which I bought last weekend!
In theory that sounds good, however in practice it doesn't work as well. OS development is a minor exercise in chaos theory. (You know because some butterfly flapped its wings in china, we have bad weather). The smallest changes can have a ripple effect through out the whole project. The more core the module. The greater the likely hood of difficultly. For me I want the whole package at once. Then you can test the whole thing and have a higher assurance of stability.
My Weblog
Why'd you put FreeBSD in the "some others" category?
-bugg
I thought of that right after I posted...duh. I'm curious though why there isn't something in place that saves the original files in tarball or something. I guess you could eventually have a lot of wasted space if you do a thing like that though.
There was a famous quote by MK about conscription that could be applied to such a product (though morphed a bit):
"Not necessarily stable, but stable if necessary."
well, storm linux (http://www.stormix.com/)is pretty much that. It's based on Debian, and has implemented a new (very graphical and very nice.. looking) installer as well as other hey-that's-pretty-neat features much like Mandrake does with Redhat.
Look out honey, 'cause I'm using technology; Ain't got time to make no apology
I've been trying to get a Debian release to download all week.
The hoops Debian make you jump through to get a version downloaded together is the biggest problem of all...
I'm trying to do the download on my company's fast comms line, but wget doesn't like my proxy, and the documentation on using wget with a proxy server is seriously minimal.
Sorry, but _I_ wasn't impressed with Debian. ALL I WANT IS THE BLOODY ISO!!!
Maybe I'll shell out the cash & buy the CD's...
OT - looking down most computer software shops in Tottenham Court Road (London, UK) this week; all of them had RedHat, lots had Suse, some had Corel and two even had Mandrake!!!
Strong data typing is for those with weak minds.
Strong data typing is for those with weak minds.
Sieg Ackermann!!!
The current Slashdot moderation system is made by gay communists!
Typical elist BSDism on display!
Bill Joy coded it in 1981, if it wasn't for that damn ATT lawsuit you linux punx wouldn't even exist!
A midrange debian, does that come with 2 doors or 4? A sunroof? I'd buy it.
The current Slashdot moderation system is made by gay communists!
-release comes from the -stable branch and 'testing' is comming from the 'unstable' branch from debian..
Then you shouldn't be downloading an ISO in the first place. If you have a slow connection, either shell out the money for a CD or a faster connection (I'd go for the latter any day), or just get:
t )/main/">
/disks-i386/current/images-1.44/rescue.bin
/disks-i386/current/images-1.44/root.bin
/disks-i386/current/base2_2.tgz (I know it's big, but don't complain!)
/disks-i386/current/drivers.tgz
:) (two were on a fast connection, so I got lazy and grabbed the entire /disks-i386/ tree)
<base href="http://http.us.debian.org/debian/dists/(dis
I think that's it... I should know, since I've done it seven times
... that's why you have Debian mirrored on SourceForge's servers :) and it's a lot more than 400 small files, last I counted. But actually it can speed up a fast connection, becuase you can have several streams coming in at once (hmmm... don't think wget can do that yet... hopefully I'll be wrong soon), so you get more share of the backbone and your local ISP's connection (especially if it's metered -- then you can max up to your metered speed, and get what you pay so much for).
(wishing I had a fast connection...)
Hmm,I don't know,what part of Testing and Unstable shall be made More explicit? I think they're quite clear.
The best way to escape from a problem is to solve it. Alan Saporta
I just installed Debian (the newly released stable version) for the first time after running Red Hat and Mandrake for the past 1.5 years. So far I'm impressed, although downloading all the packages I need is a bit annoying, but then again the control it gives you is good. Also, the Debian package managemt tool (dselect) is excellent. I had nothing but problems with RPM's, but dselect is awesome.
I'm still having some minor configuration issues, but so far I'm impressed with Debian. I'm tempted to switch to the unstable release, though. My guess is that it would be better than the stable release of Red Hat.
Anyway, the important thing is that it works, unlike RPM's.
You can make wget do that by using multiple xterms and wget's -nc (no clobber) argument. If you also use the argument to prevent it from creating a directory called ftp.whatever.com (I think it's -nH), you can download from several different ftp sites at the same time for maximum speed.
Because some of us are stuck, temporarily or permanently, on a shitty slow 56k (or worse) connection, and downloading ISOs or FTP directories just isn't practical.
This means actual CDs, mostly shrink-wrapped, which are then sold as commercial products with a version number to inform the public of the "age" of said distro.
More or less. I believe NetBSD follows more or less the same patterns?
OpenBSD's current is always more secure than release, however, and I believe also theoretically more stable as well.
OK. Let me expand a bit on my original comment.
The whole release concept seems to be a bit too limiting for Linux distros, which are composed of many different bits of software, which have completely different release schedules of their own.
It may by true that releases help maintain sanity by not breaking dependencies every other day, but, for example, potato has packages that are outdated, despite being released just a little while ago. Because some of the packages were old, I had to upgrade a few of my packages from the unstable tree, and didn't have any problems with dependencies.
Upgrading a package by a minor revision should cause no problems. Potato still has bbdb 2.00, when version 2.00.06 (even smaller than a minor version) is out and fixes an annoying bug that happens when it is used with emacs 20. Why isn't the new version in Potato?
gkrellm in Potato is still at version 0.7.4 -- I upgraded to 0.10.5 from unstable with no problem with dependencies. Potato has xmms 1.0.1, but 1.2.2 works fine with the same libraries. Can't someone just look at it, see that it works just fine and that there were no problems with dependencies, and then put them in the stable branch?
Well, that's my rant of the day. Don't get me wrong; I love Debian -- I just switched from RedHat, and found it easier to install and maintain. It just seems to me that there's still room for improvement.
To get something done, a committee should consist of no more than three persons, two of them absent.
Neither Stable Nor Unstable: A Midrange Debian?
As a nod to the famous secret agent, I think the headline for this post should have been (in Sean Connery's voice):
HAHAHA! I kill myself.
Sorry about the irrelevant post. I just had to say it.
Accountability on the heads of the powerful.
Power in the hands of the accountable.
go to www.lsl.com i think linux system labs.. the cd's are like 7 bucks with shipping. you know there are free isps with 1800 dialups and they use regular dialup networking ie they work from linux... check it out.
I dunno, not the best phraseology to use, 'eh? Makes me think of a muddy bog, not software.
Maybe "pliable" versus "crystalline" would be a better metaphor...
-subtraho
In FreeBSD:
[ Stable ]----(new code)--->[ current ]
+--------(erratas)---->[ release ]
In Debian
[ stable ] = official release + erratas
^
[ testing ] = unstable with some checks
^
[ unstable ] = new code that dinstall at least accept
^
[ experimental ] = mad science packaged -- please never let apt pointed on this one!
Fabien Niñoles - Debian Maintainer
I would use Debian if its packages weren't so ancient. Crissakes, until a few weeks ago, the OFFICIAL Debian contained Kernel 2.0!!!
A while back, Linux-Mandrake started making "a better Red Hat than Red Hat". It wasn't easy, but they got there. Making "a better Debian than Debian" would be even easier. It would be the best of Slackware combined with the best of Mandrake -- sign me up for one of those...
--
--
The testing distribution is for those who really want to do what you're suggesting- only updating packages that have survived in unstable for over a week or two. The latency is there so that the number of bugs that will be encountered by someone updating out of testing is hopefully lowered, at least relatively to one who updates out of unstable.
--Rana
I thought the whole point of Debian was that their releases weren't 'sludgey'. If Debian follow the 'sludgey' model, even if its not in their main release, what then makes them different from any other distro? The fact they use deb packaging rather than RPM?
As far as I'm concerned all this does is make Debian a slightly more 'free' RedHat.
Blaming GW Bush for the Iraq war is like blaming Ronald McDonald for the poor quality of food.
It needs to be up to date and more or less whole. If it changed that much technically you could package "unstable" and press CDs and it would still be more stable than Red Hat or others.
Respond to s
Do they plan to release CDs that can be purtchessed?
Personally I may be temporarily switching OSs to (*sob*) Win98 to get proper free internet support since no one in the linux community can give local access in Utah (too damn hot I will probably move back to Maine).
Can you get copies of a pressed CD of unstable? That would be cool. Unfortunately I don't have a good CD burner.
Respond to s
It's just creating a new category of release to make it so that others can feel better. That's all nothing so grand as creating a new red hat.
Respond to s
Really, I should know after running it for about 2 or so years until I switched to debian. The experience of sitting at my computer for 10 hours to try and find a problem that some guy didn't get because he was up since 3am Tuesday (it's Friday now) and he didn't do is anoying.
I like debian because I can actually upgrade the distribution. One of the things I don't like about Mandrake is it's use of Pentium optimizations (I run on a 486/66) and it's method of putting everything in a big unorganized directory of RPMs and it's lack of a floppy install.
Respond to s
Red Hat has a few problems. Why do they actually have so many all the time that's my question.
Respond to s
The point of this is that "unstable" will become _more_ bleeding edge, just like you want it; this will be called 'sid'. Meanwhile, 'woody' will be less bleading edge than 'sid', but still pretty darn current, and much less likely to royaly fsck things up. Sounds perfect for me.
Actually it seems that they do just as you suggest for the stable branch, at least to an extent. I know that I've run dselect in the past and had it present a bunch of minor updates to various installed packages, both common and uncommon. And the entire Debian package management system then pursues the dependencies where necessary. I think it's only minor updates that get "folded in" this way, though, otherwise they might break the QA testing on the stable release.
FreeBSD does a similar thing with the -STABLE versions. Someone once noted that each -RELEASE version stays fixed for about fifteen seconds, then it points to -STABLE.
Kind of, but FreeBSD's -RELEASE just ends up being a symlink to -STABLE very quickly. Since everything that goes into -STABLE is tested, you wouldn't really want the original released version anyway because you might be missing some important bugfixes and security updates. As someone else mentioned, this is pretty much the same system that Debian has used for its stable branch all along.
If you can't beat them, eat them.
There are (or at least, there were) efforts to use the BSD kernel inside Debian. If the list archive search weren't broken (internal server error), I could have posted a link. {-:
BTW, wouldn't there be four branches at least during the QA phase (unstable, testing, frozen, stable)?
--
this post was brought to you by Andreas Fuchs.
this post was brought to you by Andreas Fuchs.
echo [Address] | sed s/[-a-z]//g | tr A-Z a-z
Why must it be made explicit? Think about it... 99% of the newbies out there are not going to go out and acquire a copy of Debian because they are not going to know that it even exists. For them, they are safe, happy, and in the dark (like mushrooms) with whatever copy of Winblows they have, or with their iFruit. For those who do "have a clue" about these things, they are sure to realize that if it's a beta release, it's not going to be automagically the most stable system in the world. I think by now we've all been burned to some degree or another by betaware, so we should be fine.... Just my $0.02 Kierthos
Mr. Hu is not a ninja.
I understand that unstable has been rather out of date because everyone was focusing on getting Potato out. Now that it's released, I imagine unstable will get closer to the "latest", making your heart pit-a-pat a bit more.
--meredith
Sometimes a scream is better than a thesis
First of all, I'd like to once again offer translation for people who do not use debian...
Stable: Its too old to bother playing with so it won't be changing much.
I've been flamed and moderated down for mentioning this in the past but I actualy like this about Debian. It is true, a system is stable if it isn't changing much. Usualy in free software it isn't changing much becuase there isn't anything to fix, and its old enough that no one is really developing for it anyway. To proprietary distro's stable means it won't crash. But whats the use of getting the latest 4.0 release on disk when the day after you buy it a problem arises and you have to get 4.01? That isn't stable. And honestly its only the old software that no one is developing for that you can assume such stability from.
Unstable: We aren't sure where this is going yet but its the latest stuff.
Unfortunately unstable in the past has had things break. Like perl, libc, and bash. Things that darn near ruined a system and would leave to a re-install. Honestly I don't know how the bad versions got in the distro because the problems were sooo glaring and obvious that I can't assume that they were tested at all.
I've often brought up that there should be a little lag time to filter out such destructive incompetance. The response was "no, let the user beware." Really, all it would take is people running the 'incoming' distro to have things happen the way they are proposed to with the testing. But its just an automatic move from incoming to unstable, no QA whatsoever and from what they say there never will be. So this is the next most logical choice.
And just because lots of people talk about it, here is my take on the definition of Free...
Free: Doesn't threaten our guild socialism.
I personaly would like them to use 'nonthreatening' and 'threatening' instead of 'free' and 'nonfree'. It would simply be a lot more understandable. But wouldn't help keep the aloof holier-than-thou status that really attracts debian developers.
All in all if you learn their quirks in vocabulary its the best easiest funnest distro out there. And they could use your help incorporating such cool things as lothar, etc...
Open source does guarantee that bugs will be found rather than left concealed, and that they can be fixed straightforwardly. It doesn't in any sense keep them from being made or released in the first place.
The fact of the matter is that some open-source and free-software projects have a vastly better track record in terms of stability (which includes security) than some others.
This is really nice. This basically gives a 'QA' release of stable, etc, before it's really stable. woody is like a full fledged development environment, aka, things change at anyone whim. Once ready, things go into a 'QA' stage. This is a very good and tried and proven method for testing of software for stable release.
-- I'm the root of all that's evil, but you can call me cookie..
--
--
"Convictions are more dangerous enemies of truth than lies."
I've had long discussions with BSD people about what the various designations used by FreeBSD really mean. This is the what I've come to understand:
FreeBSD ~ Debian
current ~ unstable
stable ~ stable
release ~ "we stuck it on a CD"
The releases are really just snapshots of one of the other branches of FreeBSD -- basically the source tree is frozen in time and pressed onto a CD-ROM. Anyone installing from a release is urged to follow "stable" which tracks fixs to the distribution. So "release"s are really just packaged versions of the other trees, provided for convenience on CD-ROM.
Thus, it seems to me that FreeBSD doesn't really have three branches.
> hmmm..., when will the world learn?
Ick. BSD elitism.
--Lenny
I like that nobody else has to make that decision for you. I personally run unstable almost all of the time, and it works great for me, but my secretary runs stable.
Bruce
Bruce Perens.
You grab a certian mail out of the middle of a thread. And then say, "Look what they are going to do!" That is wrong! Here is the end of the mail: "So there you have it. It's coded. It works. It serves a useful purpose. I think we should use it." It does NOT say it's official policy or procedure does it?
If it's not made painfully obvious that we're dealing with two different distributions, people will start generalizing. One can forsee everything from "Debian is not user friendly" to "Debian is unstable" to "Linux suXX0rz! W1ND0ZE r001z!" Okay. I'm going overboard at the end there, but I hope my point is made.
RedHat has managed to handle this well with their RawHide release, Mandrake with its Cooker, etc etc... I don't forsee any major obstacles to Debian giving "the people what they want" with what they've described as their "sludgey" release. As long as they're kept separate, especially such that Debian doesn't sacrifice its established position with its reliable frozen releases, it's a good thing. Hell... it might tip me over the edge into installing Debian around here...
--- [DrPsycho] Coping with reality since 1975.
-DrPsycho - Coping with reality since 1975
Why not set up an el-cheapo PC running Win as a gateway box, and network your Linux machine through it. Damned counterintuitive, since it works so well the OTHER WAY AROUND (hell... forget Win-->Linux Gateway, just install Linux everywher), but at least that way you'll be able to use one of those "free access" offers.
Heh. Gets around those annoying banners too! :^)
--- [DrPsycho] Coping with reality since 1975.
-DrPsycho - Coping with reality since 1975
Unstable: Has some glaring configuration errors, is the copy I'm currently working on and is in no way guaranteed to compile.
Release: Not all features implemented, but the glaring bugs have been fixed. Will compile, has an updated readme and gets it's very own release number.
Stable: 1.0 or better. No new features added since the 1.0 release, just some minor bugfixes implemented due to other people's configurations or updated libs.
I know that these are not necessarily everyone's definitions, but they work for me.
~Hammy
"You fargin' bastiches..."
To handle dependencies.
If you update something and it loads you a new version of sopme shared library, it could hose other packages. And that *really* sucks.
By living on "unstable" releases, you can experience this first hand. And for some reason, it only happens the day before you really need to get some code written...
-
bukra fil mish mish
-
Monitor the Web, or Track your site!
Eloi, Eloi, lema sabachtani?
www.fogbound.net
Still supressing them memory of GNOME 1.0?
A deep unwavering belief is a sure sign you're missing something...
I tend to think that plenty of people would be willing to test unstable as well. Though, might it be kind of cool to allow changes to linux to be broken up into "patches" until they are incorporated into the next release. This would give people who want the latest and greatest a chance to get it without a full release or cd image. We could even allow the "patch" to be installed and save the original file contents elsewhere in case the user decides what they just installed was too unstable for their use. This way they can revert to the files they had before the "patch" was put in place.
I'm only using the word patch here for lack of a better word. There could be different types of things here...patches, add-ons, etc. I think it'd be cool though to install new software in its various stages and be able to uninstall it easily with a script.
This sort of thing may exist already and I may just not be aware of it.
Is this sort of release really neccessary? While I'm not currently running Woody(unstable now) when slink and potato were unstable, I ran those w/ only minor glitches, mostly in the package management, where something would require another package that no longer exists, but this was easily remedied by forcing the install and it almost always worked. Debian's unstable always seemed about 10 times more stable than some *ahem* officially released commercial products... The nice thing about debian is the apt utility. If you want a "sludgey" release, just don't update every day, just once a week perhaps from the unstable tree. It won't change much and most of the changes will be improvements. A new tree just seems like it will cause more work for the Debian people who(no offense) tend to move a little slowly already.
AFAIK, the [stable|unstable] does not refer to the state of the software but to the packaging that the maintainers do. Packaging is not trivial and sometimes people screw up[0]. The release cycle covers getting nifty new stuff and newer versions of traditional stuff and making sure that the packaging is not broken. At last count (right now, in fact), Potato covered 4402 packages, so it took a while.
[0] E.g.: Package foo depends on libBar being exactly v1.01, when in fact v1.01 or better will do. I've seen it, but rarely.
I've never understood why they do releases anyways. It is mostly a compilation of other peoples' software, so why not just test each indivitiual piece, along with the packages that it depends on, and if it's stable enough for public consumption, put it in the stable directory?
To get something done, a committee should consist of no more than three persons, two of them absent.
Really. I hardly ever have actual problems. In fact I wish it were *more* unsatble so that I could get upgrades of all the new actual development versions of various software like the kernel and friends. Personally having access to things like the newest lynx code in official .deb form wold be a plus. Maybe even the development versions of various GNOME apps and it's core code. Or how about updating the Xfree stuff every week like they have for the WINE code.
Respond to s
Founder's Camp
Founder's Camp
News for non-Nerds. Stuff that matters.
-current
-release
-stable
hmmm..., when will the world learn?
I noticed, when I read the thread originally, that there is some concern that this would take testers away from unstable. But I did not notice the counter-argument (which I happen to fall into). I'd love (really) to test more, except unstable is too unstable. Buggy I can handle, but not working isn't cool. I think this testing dist. would brin MORE testers in from stable.
Just my opinion, I've emailed my support and opinion, perhaps others should send their support for the idea to debian-devel.
I think this is an awesome idea, and I really hope it is formally implemented.
--
M
of open source is, that it is marked as "stable" when and only when it is really stable (and not when marketing has decided to ship the product), and yet you can still have your bleeding edge program when you like it. But dividing this into three different states is likely to confuse most people. It is better done if you obey the rule: "release often" (Eric Raymond). During testing phase you will find out which of these "unstable" releases are just "stable enough" to be used (that is, it is in the same state as most commercial software...). In Linux mailing lists, you'll see sometimes postings like "2.3.XX" is a stable version, use it if you really, really need feature YY". I think debian should use the same rulings. Just produce more developer releases, and if things turn out to be good enough, tell everyone about it (with a few warnings left). Its is better to be more specific: either, it is stable and tested and ready for the masses, or it ist bleeding edge where feature X works, feature Y is broken, and Z is useable if we have full moon. You just know what you're doing. And if you don't know, don't do it, even if it might work.
You found a sword: +4 damage, +5 moderator points