Linked by Thom Holwerda on Sun 1st Apr 2007 21:56 UTC, submitted by suka
Novell and Ximian "Nat Friedman has been one of the driving forces behind the development of the Linux desktop for a few year now. First with his own company Ximian, founded together with Mono chief architect Miguel de Icaza, after its acquisition now inside Novell. A few months ago he has been named 'Technologist of the Year' by the VarBusiness magazine for his work around the SUSE Linux Enterprise Desktop. Since then he has been promoted to Chief Technology and Strategy Officer for Open Source, besides the desktop he is also overseeing Novells server business now. During Novells Brainshare Andreas Proschofsky had the possibility to sit down with Friedman and talk about the Linux desktop, the consequences of the Microsoft agreement and the mistakes of the Hula project."
Thread beginning with comment 226431
To read all comments associated with this story, please click here.
Some Interesting Bits
by segedunum on Mon 2nd Apr 2007 00:35 UTC
segedunum
Member since:
2005-07-06

One of the interesting things though, is the work we have done closely with customers - like the Peugeot deal...we have learned a lot from that. Resulting from that there has been lot's of interoperability work, Active Directory...

I sincerely hope you're getting Peugeot to ditch AD and use a Novell client and server that will support your desktops, and Windows, under your control. Otherwise, people will continue to bitch about how difficult it is to manage Linux desktops in AD and how the obscure group policy they've just committed has no effect on those desktops. It's the chasing tail lights thing again, and all you're doing is running on a tread mill.

We also did a lot of work on Microsoft Exchange support

Sigh. Speaking of another tread mill. The only way you can support Exchange is through MAPI, and that's just too complex to implement. You'd be better off convincing people to move away from Exchange to a much more convenient, and cheaper, groupware solution that will allow Linux desktops and Outlook access on a level playing field and providing it as an all-in-one thing.

Quite what that groupware solution is, I don't know, but it certainly isn't Groupwise. What it should have though is a dead simple, no brainer migration tool to get everything out of Exchange and into the new environment inside about half an hour. Same with AD.

So all in all we've a huge 300.000-400.000 line application, we have a dozen or so engineers working on it inside Novell...

I do agree with the tone of the interviewer here. Circa 2001, Evolution was a half decent mail client, but over the years, with the addition of groupware support amongst other things, it has become an awful lot less stable.

You know, Service Pack 1 is still not out, so there is still some work left to do.

Sounds like something Microsoft might say. "That must be why we're not releasing it yet".

We have set some performance goals for those kind of issues, both the main menu and the "more applications" should open in less than a tenth of a second.

You can't just set meaningless and arbitrary performance goals, unless you're going to spend a ton of time and throw a lot of developers at the problem. You're on a hiding to nothing there.

And most of that negative sentiments don't seem to come from the people who accept patches anyway, they come from people who have a sort of "professional commentator" role in the community.

I wouldn't exactly call someone like Jeremy Allison a commentator.

Tracker on the contrary doesn't work half as good as Beagle, it doesn't do half the things that Beagle...I think what you'll see as the Beagle memory footprint gets lower and lower, it'll become even more dominant in terms of who uses it.

We keep hearing about how much less memory Beagle is going to use in the future, and indeed every Mono application. Constantly. The fact is, Tracker is written with what Gnome is natively written with and should integrate an awful lot better with the desktop as a result. They look as if they're going to be collaborating with Nepomuk and through FD as well, which is great.

Novell are going to be stuck maintaining an open source application in-house while Gnome and KDE move ahead with better integration of desktop search, right within their infrastructure. Better integration of search in the desktop is the way to go, rather than having some entirely separate, polling service dragging down your system that has to react to what you do.

Are you interested in doing something like Time Machine / Shadow Copy...Yeah, I think that's interesting, it could be very good to have something like that, I'm not sure exactly how to do that.

Volume Shadow Copy could be done easily with LVM, but something like time machine would need something more fundamental at the filesystem level - certainly if it was real time. However, you might be able to do something like it with better integration with something like Bacula, and you'd get networked backup as well!

One thing I commented on a while ago was that Novell look as if their desktop activities are going to consume more and more resources, and require more and more in-house developers - hacking on open source software whose functionality they are largely duplicating. I can't imagine that being a good thing.

Edited 2007-04-02 00:38

Reply Score: 5

RE: Some Interesting Bits
by jpobst on Mon 2nd Apr 2007 00:52 in reply to "Some Interesting Bits"
jpobst Member since:
2006-09-26

I wouldn't exactly call someone like Jeremy Allison a commentator.


I wouldn't exactly call your example of one person as 'people'.

Reply Parent Score: 5

RE[2]: Some Interesting Bits
by segedunum on Mon 2nd Apr 2007 20:01 in reply to "RE: Some Interesting Bits"
segedunum Member since:
2005-07-06

I wouldn't exactly call your example of one person as 'people'.

Oh, you wanted a list? I merely used Jeremy as an example, considering he used to be employed by Novell. I'm sure you can find an awful lot of people contributing to an awful lot of quite important open source projects who thought the Novell/Microsoft deal was a seriously bad idea. There's was too many people everywhere commentating on it to get a full list of names. Quite how he claims to know this, I don't know.

Unfortunately for him, Nat Friedman trying to dismiss the whole thing as a lot of fuss from commentators doesn't make the deal Novell made any less ridiculous, stupid, any less of a legal circumvention of the GPL by a supposed open source company or any less damaging to the perception of open source software. Which is the real point.

Reply Parent Score: 5

RE: Some Interesting Bits
by sbergman27 on Mon 2nd Apr 2007 04:33 in reply to "Some Interesting Bits"
sbergman27 Member since:
2005-07-24

"""
We keep hearing about how much less memory Beagle is going to use in the future, and indeed every Mono application. Constantly.
"""

Yes. They do keep flogging that dead horse, don't they?

Beagle is often held up as a shining example of a great Mono app.

But when one looks at all closely, Beagle starts to look like a good argument against it.

Beagle is nice. It really is.

But the memory requirements are pretty embarrassing.

More discouraging than that, though, is the slow progress compared to Tracker.

From a late start, Tracker has come from way behind and has mostly caught up with Beagle. And it is actually a much more ambitious project.

On the other hand, Beagle had the core of their code handed to them on a silver platter in the form of the Apache Lucene project[1], a Java indexing framework which they simply ported to C# and *Bam!*, they had a "Mono" indexing engine.

What they actually wrote, using the awesome development power of Mono, was the crawler that feeds the data to their Lucene port. Just the crawler itself, though, and not the programs that actually convert the various documents to text. Those are third party and largely written in C, I believe.

The other piece that they had to write was the Gnome applet to interact with the user.

Oh, and the scheduler for the crawler which is supposed to let Beagled do its work without impacting yours. (And is the most disappointing part of Beagle, IMO.)

It really makes one wonder *why it took them so long* to put a few third party apps together with a little glue code written in C#.

The *concept* of Beagle is neat.

But far from being a shining example of the power of Mono, upon closer inpection, Beagle is actually more of a warning beacon to the rest of us.


[1] http://en.wikipedia.org/wiki/Lucene

Edited 2007-04-02 04:50

Reply Parent Score: 5

Cross platform
by s_groening on Mon 2nd Apr 2007 10:41 in reply to "RE: Some Interesting Bits"
s_groening Member since:
2005-12-13

I agree with you!

But apart from your points, what I miss from Beagle with it being a Mono app, is a more standalone character. I'd like for Beagle to be more cross platform so that I could choose to use features like live queries on e.g. Open Solaris or BSD as well as Linux, but since that part i all tied to the libinotify kernel part ...

I think those sorts of aspects of a project needs to be taken into consideration, since the open source community always touts the freedom of choice part quite strongly. Plus it has always been the thought that choice and alternatives are what makes better programs which in turn also counts for operating systems.

Personally I feel that one should be able to choose ones (major open source) OS of choice and still be able to benefit from development efforts concerning major projects like e.g. Gnome - and that's why I'd like for Tracker to slip past Beagle to offer a true freedom of choice!

But the idea of the open source 'Spotlight clone' is great, it could just be even greater...

Reply Parent Score: 2

RE[2]: Some Interesting Bits
by segedunum on Mon 2nd Apr 2007 20:06 in reply to "RE: Some Interesting Bits"
segedunum Member since:
2005-07-06

Beagle is often held up as a shining example of a great Mono app.

But it isn't really a Mono app. It's just a rewrite of Lucene in C#, which is the most crucial part about it. Given the overhead Beagle clearly has I have just never seen the justification for writing it with Mono. It's not spectacular merely by it having been written in the ultra cool, language to end all languages, C#.

Reply Parent Score: 5

RE: Some Interesting Bits
by kelvin on Mon 2nd Apr 2007 10:43 in reply to "Some Interesting Bits"
kelvin Member since:
2005-07-06

Circa 2001, Evolution was a half decent mail client, but over the years, with the addition of groupware support amongst other things, it has become an awful lot less stable.

Evolution has had groupware support since its inception. The project was, in fact, started in order to create an "Outlook killer"; opinions differ on whether or not it has succeeded in accomplishing that goal.

Reply Parent Score: 2

RE: Some Interesting Bits
by abraxas on Mon 2nd Apr 2007 14:26 in reply to "Some Interesting Bits"
abraxas Member since:
2005-07-07

Quite what that groupware solution is, I don't know, but it certainly isn't Groupwise. What it should have though is a dead simple, no brainer migration tool to get everything out of Exchange and into the new environment inside about half an hour. Same with AD.

Half an hour!? I wish any mail migration took that little time. It's simply not feasible in most situations considering the amount of data that has to be processed.

I do agree with the tone of the interviewer here. Circa 2001, Evolution was a half decent mail client, but over the years, with the addition of groupware support amongst other things, it has become an awful lot less stable.

Evolution went through a rough spot in the early version 2 series but is very stable now. I use it on a daily basis and seldom have a problem with it.

We keep hearing about how much less memory Beagle is going to use in the future, and indeed every Mono application. Constantly. The fact is, Tracker is written with what Gnome is natively written with and should integrate an awful lot better with the desktop as a result. They look as if they're going to be collaborating with Nepomuk and through FD as well, which is great.

Mono does take a decent amount of resources. Beagle in particular takes a large chunk of memory but it is perfectly usable on a modern machine. For me searches are instantaneous. There is nothing slow about Beagle unless you are on a memory constrained system.

The last time I looked Tracker still did not have nearly as many features as Beagle but most of them were being worked on by the Tracker team. I use Beagle now and haven't tried Tracker yet but I don't see a need to unless I want to go backwards in functionality. Does anyone know if Tracker actually has a stable version that competes with Beagle feature for feature yet?

Reply Parent Score: 2

RE[2]: Some Interesting Bits
by Jamie on Mon 2nd Apr 2007 15:04 in reply to "RE: Some Interesting Bits"
Jamie Member since:
2005-07-06

Tracker is a lot lot faster and lighter than beagle and does not slow your machine down while indexing.

Search time average is 20 milliseconds and we can index over 100 text files per second - you wont find a (significantly) faster indexer.

We have had reports of tracker running well on a 700MHz pentium III with 128 MB ram so its by far the better system if you want to run it on all kinds of systems

Its architecturally superior to Lucene (tracker is a differential indexer so can handle changes to files better than Lucene which must delete and reindex after every minor change) and is better optimised for desktop usage.

AFAIK its the only indexer with an RDF triple store integrated so is the only one that can currently deliver an integrated desktop solution that spans all metadata. When Beagle has a triple store then we can compare features with tracker.

The notion that tracker is way behind beagle is simply not true.

We do virtually all the same file formats beagle does.
We do Evolution Emails (in svn)
We do Gaim Logs (in svn shortly)
We do Application Files

We will be adding our external services support over Easter which will allow third party apps to act as indexers for tracker (and we have patches for Tomboy notes and Liferea already done - they just need to be integrated)

Also there's a GNOME SoC to make Epiphany use tracker to store all its bookmarks and history.

So i think we reaching a point where we are very close to what Beagle does. On the metdata database front we are way ahead of them as we have:

1) highly optimised triple store
2) Ability to cross query indexer and triple store
3) RDF Query implementation for powerful searches
4) No duplication of metadata in triple store and indexer (unlike Lucene our indexer does not store any metadata and is metadata agnostic so can scale better)

Reply Parent Score: 5

RE: Some Interesting Bits
by natfriedman on Wed 4th Apr 2007 22:40 in reply to "Some Interesting Bits"
natfriedman Member since:
2007-04-04


I sincerely hope you're getting Peugeot to ditch AD


We would definitely like that to happen (and are trying to convince Peugeot of the same), but if we make adoption of the Linux desktop contingent upon people substituting out all the rest of their IT systems, then it will never happen. So interoperability is important as we phase Linux desktops into companies.


I wouldn't exactly call someone like Jeremy Allison a commentator.

Jeremy is a superb hacker and a friend and I definitely was not referring to him.

Reply Parent Score: 1