history of CADT

Posted May 4, 2016 0:19 UTC (Wed) by tialaramex (subscriber, #21167)
In reply to: history of CADT by louie
Parent article: Devuan Jessie beta released

_Everybody_ who sets out to burn things down and start from scratch convinces themselves that _this_ time they're doing it for the right reasons. Perhaps some of them are even correct. But certainly most of them are wrong, yet no less convinced anyway. I have worked on several "burn it down and start from scratch" projects, some of which are GPL'd and some not, and the only one that I'm even somewhat convinced was appropriate after decades of experience was the one which most resembles Brooks' "build one to throw away" pilot system approach.

In particular your apparent belief that if roughly the same group of people write a program again it will have completely different and unrelated bugs is pure wishful thinking. Those people are going to keep making the same mistakes, they're going to keep forgetting the same basic use cases, keep testing only the scenarios that apply to them and not other scenarios, and keep tripping over the same misunderstandings of their language, their libraries, and their environment. But now instead of ten unread bug reports over five years saying that app X can't do Y properly, you can start "fresh" and discover a remarkable thing - app X version 2 can't do Y properly either. Huh, who'd have thought? Only everybody who has been paying attention :/

to post comments

history of CADT

Posted May 4, 2016 5:58 UTC (Wed) by louie (guest, #3285) [Link] (25 responses)

I agree with your general skepticism of rewrites - they're often done by inexperienced developers for bad reasons, and (as with most software) often take longer than they planned. But they're also often done for good reasons, and they're almost always criticized by people who have no idea about the underlying user needs or suitability of the technology.

I think in the case of GNOME, history bears out that we were generally right to make the decisions we did. Projects tried to fork what was left of GNOME 1, and went nowhere; they found that both the old infrastructure was hard to modernize (i18n, a11y, etc.) and that there wasn't actually much interest despite some very loud complainers. (Again, sounds familiar!)

The main fork of GNOME 2 (MATE) has done somewhat better - there are a lot more happy MATE users than there ever were of GNOME 1, which is a great credit to the MATE developers! Still, Fedora and Debian have kept GNOME as their default, and debian popcon suggests 6x as many people have GNOME 3 installed as MATE (based on the installs of the respective file and window managers). In terms of developer activity, the 20th most-active MATE module on github was modified 26 days ago; the 20th most-active module on GNOME 3 was modified 9 hours ago. (You have to go through 240+ projects to find the GNOME module modified more than 26 days ago.) And the 600+ extensions in extensions.gnome.org shows that the decision to move to a scriptable shell interface (the biggest "rewrite" of GNOME 3) is both drawing developer interest and empowering users. So, yes, perhaps incremental improvement of GNOME 2 could have been successful, but a group of folks are literally trying that and not doing well by most reasonable metrics as the main core contributors.

history of CADT

Posted May 4, 2016 6:46 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link] (11 responses)

> I think in the case of GNOME, history bears out that we were generally right to make the decisions we did.
You guys _REALLY_ need a reality check.

history of CADT

Posted May 4, 2016 7:33 UTC (Wed) by louie (guest, #3285) [Link] (10 responses)

Because the old code bases (or similar ones) are doing so well, or...? I mean, obviously we're not Android or iOS, but if that's your standard for success pretty much all open source is a failure. So what's the alternative path we should have followed? Would we somehow have become Android if only we'd stuck with GTK 1? Would we somehow have become OS X if only we'd kept our 1.4 preferences dialog? I think it's pretty clear that, even if they didn't take over the whole world, the big gambles we made were a lot better than sticking with what we already had. The one metric that suggests otherwise is application uptake: maybe if we'd stuck with GTK 1 we'd have more apps? (We certainly seemed to have more Back In The Day.) But I suspect even there, the answer was that we needed to be more bold and do more rewriting (mono or Java), not stick with the old stuff.

history of CADT

Posted May 4, 2016 8:00 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link]

> Because the old code bases (or similar ones) are doing so well, or...?
I'm talking about achieving your goals.

From what I see, there are now more desktops running Unity than GNOME3. And even RedHat ships GNOME3 with Classic mode. The fact that people are still using (and actively developing!) MATE built on GTK2 released back in 2002 speaks volumes.

You've also lost applications. Many projects actually switched from GTK2 to QT rather than to GTK3: https://blog.wireshark.org/2013/10/switching-to-qt/ , https://subsurface-divelog.org/ , Unity, http://wiki.lxde.org/en/LXDE-Qt and so on.

And it doesn't appear that all that pain has attracted new users in great numbers - desktop Linux marketshare still hovers around 1%.

history of CADT

Posted May 4, 2016 23:12 UTC (Wed) by bronson (guest, #4806) [Link]

> But I suspect even there, the answer was that we needed to be more bold and do more rewriting

I have to ask... Do you really think Gnome's problem is not enough rewrites?

I'd guess that would have encouraged me to leave earlier. Even less stability doesn't seem like a winning strategy to me. But, admittedly, I'm not in Gnome's target audience.

history of CADT

Posted May 10, 2016 8:15 UTC (Tue) by paulj (subscriber, #341) [Link] (7 responses)

Your 1.4 prefs dialogue is using the stock widget look. If you were trying to make 1.4 look bad, that's what you'd choose. I doubt anyone shipped with stock/no-theme as the default theme. There were a lot prettier themes. E.g. "ThinIce" still looks fairly good today.

history of CADT

Posted May 10, 2016 15:37 UTC (Tue) by raven667 (subscriber, #5198) [Link] (6 responses)

That is a very weird thing to say, that showing the default operation of a particular software would be considered "trying to make it look bad" and that this would be considered a defense of the software in question. Are you saying that the people who built this software and chose the default look were "trying to make it look bad" ??!! My head goes all 'splody trying to understand what you are implying here.

history of CADT

Posted May 10, 2016 15:46 UTC (Tue) by paulj (subscriber, #341) [Link] (5 responses)

Louie seemed to be arguing that the GTK look from GNOME 1.4 days was ugly and using that as evidence for some greater argument about need for GNOME to rewrite stuff. I agree the /default/ GTK library look from those days, as shown in the example, was indeed a bit ugly. However, I don't think any distro shipped with the stock/no-theme default widget set as the distro-default. I don't know why the library default was ugly, but that library default look was *not* what users got anyway in those days - so using a screenshot of that library-default seems unrepresentative, and hence not a good support for whatever argument louie was making.

history of CADT

Posted May 10, 2016 16:00 UTC (Tue) by rahulsundaram (subscriber, #21946) [Link]

> Louie seemed to be arguing that the GTK look from GNOME 1.4 days was ugly

That isn't it at all. The image shown is the control center setting overloaded with geeky options. The look of the toolkit is irrelevant to the point being made there.

history of CADT

Posted May 10, 2016 16:11 UTC (Tue) by pizza (subscriber, #46) [Link] (1 responses)

I think the point wasn't the actual widget appearance, but the complexity of the preference dialog and the implications it had for both usability and testability of GNOME 1.x as a whole.

history of CADT

Posted May 10, 2016 16:36 UTC (Tue) by paulj (subscriber, #341) [Link]

Ah. :)

Ok. On that, I'd whole-heartedly agree. GNOME 2 was a huge improvement in that regard. It should be noted that, a factor in the (eventual) success of GNOME 2 was that the changes were driven by empirical HCI studies carried out by Sun Microsystems. The results of which led to the GNOME 2 HIG.

GNOME 2 wasn't just arbitrary change. It was change based on objective evidence.

history of CADT

Posted May 10, 2016 16:12 UTC (Tue) by raven667 (subscriber, #5198) [Link] (1 responses)

> Louie seemed to be arguing that the GTK look from GNOME 1.4 days was ugly

I didn't get that at all, I didn't see this as making a comment about the aesthetic qualities of grey, which I think is largely irrelevant, but as a comment about the organization and number of preferences, which was greatly simplified in the move to GNOME 2, so that every user didn't have to wade through a cacophony of irrelevant sliders and tabs, to change the few options that were most likely to be changed, while still retaining many of the variables exposed in a more advanced interface for those who want it.

This was a major change and rewrite that most people seemed to like better than the old tool in time, you could make the same kind of comparison between Sawmill and Metacity, would GNOME be stronger today if Sawmill had stayed as the default, to maintain compatibility, same as if the original Control Center had stayed?

history of CADT

Posted May 10, 2016 21:41 UTC (Tue) by nix (subscriber, #2304) [Link]

I would certainly be hugely happy if Saw{mill,fish} had stayed alive, rather than its main developer being hired by Apple on the condition he stopped working on it. It was and is the best Lispy WM out there, bar none.

history of CADT

Posted May 4, 2016 6:46 UTC (Wed) by louie (guest, #3285) [Link]

Sorry, not doing *as* well. They're definitely doing pretty well overall, and again - serious kudos to them. I'm glad they're using the freedom our licenses granted them and successfully empowering users that way.

history of CADT

Posted May 4, 2016 7:48 UTC (Wed) by johannbg (guest, #65743) [Link] (10 responses)

"600+ extensions in extensions.gnome.org"

Does windows have "extensions"
Does MacOS X have "extensions"
Does Android have "extensions"

It goes without saying desktop environment should be growing application not extension to workaround it's poor design.

+If memory serves me correct then extensions were never planned to be part of Gnome but grew out of necessity to their keep end user base.

Now the fundamental problem with Gnome is the instability of it's UI design and always has been ( 1.x, 2.x 3.x does not matter ).

That was the thing novices end users complained most to me about when I was Fedora Ambassador.

Those end user did not what shiny new things with "effects" and gazillion default applications to be confusion about.

They wanted a boring stable UI with the exact same applications in the exact same place as before so they did not have re-learn where applications and system configuration was placed or start using new applications they where entirely unfamiliar with.

And that was why novice end users did not use Fedora which ironically is to many ( Red Hat ) people it's "target audience" and why the Red Hat desktop team is forcing Fedora's release cycle in sync with the Gnome one again ( despite 10 years of history telling them not to but hey let's tear the community a new one with now with no mass rebuilds to achieve that impossible goal! Fracking idiots )

history of CADT

Posted May 4, 2016 10:53 UTC (Wed) by pizza (subscriber, #46) [Link] (8 responses)

> Does windows have "extensions"
> Does MacOS X have "extensions"
> Does Android have "extensions"

FYI -- Yes, Yes, and Yes.

> +If memory serves me correct then extensions were never planned to be part of Gnome but grew out of necessity to their keep end user base.

Extensions were *always* part of the plan and were a core feature of Gnome-Shell. Specific extensions, on the other hand, may not have been. Which leads me to...

> It goes without saying desktop environment should be growing application not extension to workaround it's poor design.

No, it's called designing sufficient flexibility so that functionality use cases not part of the core design can be added by those who want or need said functionality. There's quite a lot of stuff that doesn't rise to the level of an "application" or by necessity needs information only known to the shell itself.

> They wanted a boring stable UI with the exact same applications in the exact same place as before so they did not have re-learn where applications and system configuration was placed or start using new applications they where entirely unfamiliar with.

Users want nothing to change except for the things they want to change, then loudly complain when the changes they want require changes elsewhere.

Meanwhile, if you wanted a "Boring stable UI" then we'd all still be using CDE. Heck, one of the "enterprise" applications I have to use would look right at home on that platform. It's also a festering pile of swill.

history of CADT

Posted May 4, 2016 11:09 UTC (Wed) by johannbg (guest, #65743) [Link] (7 responses)

> Does windows have "extensions"
> Does MacOS X have "extensions"
> Does Android have "extensions"

"FYI -- Yes, Yes, and Yes."

Please provide me with links to documentation to the extension framework those other OS provide.

history of CADT

Posted May 4, 2016 11:24 UTC (Wed) by pizza (subscriber, #46) [Link] (6 responses)

> Please provide me with links to documentation to the extension framework those other OS provide.

Okay, I'll do this, even though what you're asking is literally as simple as plugging "X shell extensions" into google and pressing "I'm feeling lucky".

"windows shell extensions" returns: "creating shell extension handlers" on MSDN:

https://msdn.microsoft.com/en-us/library/windows/desktop/...

"Finder shell extensions" returns "App extension programming guidelines" on Apple's developer site:

https://developer.apple.com/library/ios/documentation/Gen...

Now Android's a little more complicated; "extensions" in the classical sense are really a per-application thing, and there are many applications which implement their own extension mechanism. However, the Android core is built around the concept of "intents" which allow arbitrary code to implement various activities. For example, every application in the "Send to..." list is there because it's registered an intent that can be used by anyone). Here's the "I'm feeling lucky" link returned from searching for "android intents", called "Intents and Intent filters" on the android developer site:

http://developer.android.com/guide/components/intents-fil...

history of CADT

Posted May 4, 2016 14:10 UTC (Wed) by johannbg (guest, #65743) [Link] (4 responses)

"Okay, I'll do this, even though what you're asking is literally as simple as plugging "X shell extensions" into google and pressing "I'm feeling lucky"."

You where the one that was claiming that the other OS had extensions and it's framework and their implementation was comparable with Gnome so it goes without saying you refer to what exactly you yourself are referring to so one can make the same comparison and reach the same or different conclusion as you did.

I view extension or "plugins" as highlighting underlying design deficiency ( the more extension people install or the higher usage of specific extension indicate something that should be the part of the default design since from my point of view people should not have to install extension or plugins to get the functionality they require from the desktop environment but providing the framework to overcome limitation of the desktop environment is not something I'm against ) so our opinion on that topic differ.

And the boring stable UI conclusion comes from being Fedora Ambassador for 8 years installing Fedora on those novice end users hardware and helping them with problems they experienced running and using Fedora as their day to day desktop in the process. These where regular technology challenged individual not tech savy people who know or want spending hours setting up, tweak (or otherwise fight the desktop environment to be able to use it.

In the end of the day what matters the most is what works for the end users, ( them, they are the ones who will be using that computer on a day to day bases not me, not you or anyone else which often seems to be forgotten ) and makes their life easier and helps them get their work done and is least in their way in doing that.

If Microsoft Windows works for them good, if OS X works for them great, if Linux works for them fantastic but in reality Linux on desktop is failing to work for 98.35% [1] people using computers as their desktop on the planet that's undisputed fact otherwise it would be more popular and would be more widely used.

1. https://www.netmarketshare.com/operating-system-market-sh...

history of CADT

Posted May 4, 2016 15:14 UTC (Wed) by anselm (subscriber, #2796) [Link] (1 responses)

I view extension or "plugins" as highlighting underlying design deficiency ( the more extension people install or the higher usage of specific extension indicate something that should be the part of the default design since from my point of view people should not have to install extension or plugins to get the functionality they require from the desktop environment but providing the framework to overcome limitation of the desktop environment is not something I'm against ) so our opinion on that topic differ.

Sometimes it is difficult to figure out before the fact what functionality people would actually like to have. In that sense, providing an extension framework is like sowing a huge lawn in the town square and then later paving over those paths where people have been walking a lot, rather than paving a bunch of footpaths first and hoping that people will follow them and not walk on the grass. Similarly, the most popular extensions can then be made part of the core product.

It would be great if a large and complex piece of software like GNOME could be all things to all people from the get-go, but that's not what usually happens. In view of this, an extension framework is the next-best thing because it lets users scratch their itches in all the places that the original designers didn't foresee or didn't consider important. On the other hand, deliberately omitting basic functionality that is obviously desirable to a large proportion of known users, and relying on extension developers to put it back (which I hear is something the GNOME developers subscribe to – I'm not a GNOME user myself so don't have first-hand experience) is probably not the wisest approach.

history of CADT

Posted May 4, 2016 17:49 UTC (Wed) by rahulsundaram (subscriber, #21946) [Link]

> Sometimes it is difficult to figure out before the fact what functionality people would actually like to have

It also makes niche but very useful extensions possible. The existence of extension frameworks is a good thing.

history of CADT

Posted May 4, 2016 15:34 UTC (Wed) by anselm (subscriber, #2796) [Link]

but in reality Linux on desktop is failing to work for 98.35% people using computers as their desktop on the planet that's undisputed fact

I don't think it is quite correct to say that desktop Linux is “failing to work” for all those people when the reasonable assumption is that the vast majority of them has never actually tried a Linux system in the first place. Something “failing to work”, in my opinion, implies a reasonable attempt to make that something work to begin with. (Incidentally, arguably there are many people whose entire computer needs could be adequately covered by Firefox and LibreOffice, and it would probably not matter in the least to them whether these run on Windows or Linux.)

As long as virtually all desktop PCs come with a pre-installed operating system that isn't Linux, and only comparatively few people actually care enough to install Linux, talking about Linux “failing to work” for everyone who stays with their factory-provided (non-Linux) default operating system is unreasonable. In any real sense, Linux has only “failed to work” for those people who actively installed it, gave it a try, and then went back to whatever they used before, and that will be considerably fewer than 98.35% of all desktop PC users.

history of CADT

Posted May 4, 2016 15:51 UTC (Wed) by pizza (subscriber, #46) [Link]

> In the end of the day what matters the most is what works for the end users, ( them, they are the ones who will be using that computer on a day to day bases not me, not you or anyone else which often seems to be forgotten ) and makes their life easier and helps them get their work done and is least in their way in doing that.

End-users want absolutely nothing to change. Except for the things they want changed. And then it needs to be exactly the same, only different.

...Yes, I've been through this quite a few times. Suffice it to say that I have very strong feelings on the subject.

> If Microsoft Windows works for them good, if OS X works for them great, if Linux works for them fantastic but in reality Linux on desktop is failing to work for 98.35% [1] people using computers as their desktop on the planet that's undisputed fact otherwise it would be more popular and would be more widely used.

I don't doubt that Desktop Linux is insignificant (Steam claims ~1% of their users do so on Linux, for example) but that Netmarketshare.com's claim that *Windows 3.1* comprises of 0.45% of the installed base makes me distrustful of the quality of their other figures.

I also wonder where they're slotting in Chromebooks, which on their own would account for more than Linux's total reported share.

history of CADT

Posted May 4, 2016 17:28 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link]

> "windows shell extensions" returns: "creating shell extension handlers" on MSDN
Shell extensions are not really comparable with GNOME plugins. They were used to do stuff like virtual filesystems in Explorer or additional toolbars, not to change the way Windows looked.

Now, since Windows tried to preserve binary compatibility, lots of products hooked into the deep levels of system and provided lots of customizations.

Mac OS X had even less customizability.

history of CADT

Posted May 4, 2016 11:49 UTC (Wed) by nye (subscriber, #51576) [Link]

>Does windows have "extensions"

Yes, in unimaginable numbers. First literally - shell extensions are extremely common, and a great many applications come with them.

Also less literally there are a huge number of utilities which aren't technically 'shell extensions' but which are functionally the moral equivalent in that they are programs that adjust the behaviour or functionality of bits of Windows. For example, simple utilities like 7+ Taskbar Tweaker do the kind of thing that might be done by a Gnome extension, and you might also consider larger scale things like Classic Start Menu or DisplayFusion to be comparable.

history of CADT

Posted May 5, 2016 18:52 UTC (Thu) by sionescu (subscriber, #59410) [Link]

Having to install 16 extensions to make gnome-shell useable, that are sometimes incompatible even among themselves and break with updates, that does not "empower" me, just makes me curse at GNOME devs for not having those features in the base distro.