Page 2 - Open-source needs a game plan...

I agree that there is a lot of excess slag out there in open source development. Browsing through SourceForge and Freshmeat, you find a lot of underplanned, underdeveloped, non-mature software. And, it's true that there are weak areas in typical O-S development that could use some more work.

But, I think there is a big difference between the typical small O-S project, and the few biggies that are undeniably successful. There is plenty of focus in that market.

And yes, open source focuses more on the back-end than on the GUI stuff. Part of the territory, I'm afraid. Techies tend to be more interested in the logical capabilities of their software than in pleasing the typical computer '(l)user'.

But, this is where ideas like Mozilla/XUL actually come in, IMHO. It provides a capability for logical manipulation of your interface without spending all that fluff time worrying about your GUI implementation on this or that platform.

Those idiots who mention that Netscape/Mozilla is 'not a quality product' and 'it's very difficult to talk to anyone' have obviously never used IE. (Which one has had more security flaws? Broken standards? Performance degradation over time?) And they probably never even sent one email to the dev newsgroups asking for help. Ask jpenn; he already has a couple highly placed Mozilla developers corresponding with him regularly.

The fact is that no commercial software house has even attempted something on the scale of the Mozilla project. It's development is on a par with the innovation in Apache 2, which again is something no commercial entity has even attempted. (I say "foo" to those who argue that open source never innovates, but just imitates . Just look at the actual capabilities of Apache 2 and Mozilla.)

Mozilla is not just a browser; it is an application framework that is completely cross-platform, cross-language and scriptable, and where the same GUI elements can be delivered by a) local binary application b) local XML/javascript, and c) remote HTTP. Find me any comparable environment by a commercial concern. There is Java, which is cross-platform, but compiled rather than scripted, and is accessible by one programming language only. We have .NET, which is scriptable, deliverable by HTTP, etc... but is not cross-platform -- yet (guess what? the OSI guys are handling that end also...)

And Mozilla may not be completely production-ready for the same reason as Apache 2 is still not really production quality: they are going into uncharted territory, and trying to do it right! That takes time and serious work. So, I don't think the problem with Netscape/Mozilla is a lack of focus. I think it is just that the developers took on a very ambitious project, and are succeeding, but needing a longer time frame than originally estimated. Most of Microsoft's "good" products are much older than Mozilla, and they have much narrower requirements.

In general, though, I have a couple of opinions as to ideal placement of O-S work and proprietary commercial work in the computer market: it seems to me that open source is the best way to pursue things like Mozilla, Apache, etc..., while proprietary commercial concerns are sometimes better equipped to handle specific projects such as groupware/business/accounting/engineering, etc... ("Vertical markets", to be buzzword-compliant). You even see this happening in OSI-based companies such as Suse and RedHat.

In other words, O-S stuff is more viable to occupy certain parts of the computing spectrum, while commercial software may be more viable for other parts. I see open source as perfect for broad, generic things like webservers, application platforms, operating systems, browsers, programming languages, protocols, etc... but for certain specific business needs, where the requirements are very specific and exhaustive (and possibly irritating to the techie), proprietary software tends to provide a more user-friendly gift-wrapped kind of solution. But then again, when you move on down the spectrum to smaller, perhaps more quirky, fun kinds of software, such as photo editors, games, etc... open source again competes viably, because of the reduced requirements, and the "fun" factor. (I.E. software that scratches a personal itch ).

The good thing, in the above case, is that businesses can save themselves and the consumers a LOT of money by using open source components to build their proprietary vertical software. I don't see this as a bad thing at all, as long as the license allows for it (BSD, MPL, LGPL, etc...). Any company seriously using open source components gives back in the form of quality enhancements to the O-S components, etc... because it is simple enlightened self-interest (see if you can count the reasons). You can't show me a successful company using open source components that doesn't give back (with the possible exception of Microsoft's use of BSD components-- but even then, Microsoft is helping in the port of .NET to FreeBSD).


Maybe, but I have to admit that it is fun--and breathtaking--to see the scope of all the crazy ideas that have been tried with software. Give people a little freedom, and see the fun commence. We shouldn't see this whole thing as a battle between open source and commercial, but rather a search for the best solution to each need. I, for one, love the fact that there is so much range, lack of focus, etc... to the open source world. I hate the idea that everything should be reduced down to the most standardized, unified representation.

For this same reason, I hate the arguments out there (from experts like Jeff Ranking, among others), who say that every single GUI should look and work exactly the same way, so there would be no surprises moving from one desktop to another, blah, blah... Rankin even says that the user shouldn't be able to customize the desktop colors or fonts. Sure, take all the fun out of computers. Let's just make the world a cold, grey place, why don't we? Let's just assume that every computer user is a complete idiot and can't possibly handle even the slightest difference in how things work. Every user is "grandma", and is afraid the computer is broken when a window closes.

No, personally, I don't think there is a way to make the open source world standardize and unify too much more. They are already much more unified than the various commercial vendors. Ever seen a Windows user try to understand Solaris? Moving from Gnome to KDE makes much more sense. Moving from MySQL to PostgreSQL is much easier than moving from SQL Server to Oracle. Moving from PHP to Perl is easier than VB to Java. You get my drift? In this case, the fault is not with open source but with the fact that the computer world is already quite fragmented. It only seems unified to people who only use Microsoft. And even then, Microsoft isn't completely unified with Microsoft!! Just try to make sense of all the APIs.

Some of the things I wish would happen are out of pure selfishness. I want Linux to succeed on the desktop, so that the places I work at will start using it more and I'll get to spend more time working with it. I also would like to see more parity on the desktop where a user or company can pick any machine they like.

The reason I think desktop computers should behave consistently isn't because I want to stifle creativity or uniqueness. That should be part of the output or content of the program. But the way a user operates the program should be intuitive. A user should KNOW what's going to happen when he clicks a mouse or hits a key. Programs don't have to do the same things or look exactly the same but they should behave in a consistent manner. It would be like automobile manufacturers putting the clutch, brake and accelerator in a different order. I think it would cause problems.

Just because I suggest that it might be a good idea to focus efforts on a set of core programs doesn't mean an individual wouldn't be free to try anything they like. I view it it more as a concerted effort to gain a greater acceptance of open source programs and operating systems. Think of it as a big stick of dynamite to blow open a logjam (aka monopoly) so things can flow more smoothly and evenly. As much as I like Linux, I've gotten a lot of good use and work out of Microsoft products. I don't hate them or their products. I think their lopsided control of the desktop is a problem and we have the power to correct it.

There is a valid argument for not doing any of this: We like being quirky. We want to have a geek operating system that the average person is incapable of using. Like an exclusive club that is only inhabited by people who are willing to dig into some code, we don't want just anybody to say they are a Linux user.

You see, I don't think its an either-or problem. Most Windows users have no trouble understanding a KDE desktop. Many Windows users I have given a Mandrake or RedHat disk had no trouble doing a complete install. OpenOffice is different than Microsoft Office, but no more difficult to install or learn.

My daughter is just turning 3, and just for fun, I gave her an account on one of my Linux systems, and let her play with the desktop all she wants. She has managed to find most of the things that interest her, such as the games, TuxPaint, etc... with no more trouble than she had with my wife's Windows system.

My point is that the situation is more complex than most people make it out to be. And I don't think it's just a question of useability, or software robustness. I think it's also in large part a question of perception, and media hype, as well as other issues such as (a big one here) lack of vendor-provided drivers for hardware.

The classic "Which came first: the chicken or the egg."

Vendors won't take the time to produce the drivers until there's a demand from the users. The users won't use Linux until they have drivers for their hardware. Certain developers won't consider Linux as a target platform until there are a significant number of users. etc. etc...

My suggestions earlier in this post were put forth to try and get a core set of applications developed that would be of interest to a wide-spread group of users (schools, governments) and wouldn't need special hardware.

A single Linux Apache server can service 1000's of users. Its also very transparent. The user has no awareness of what OS is serving the information to his browser. A desktop is typically used by just a few people. At my office our desktops outnumber our servers ~10:1. If all the desktops were using Linux then there are a lot more instances of Linux running and the OS is more "in the user's face" - they're now aware of it. The point of this is simply that when Linux becomes more viable on the desktop, there will be an explosive growth in the number of copies running.

The situation is complex but I'm not thinking in terms of next week. I'm trying to look a few years down the road. A few years ago, there were still Windows 3.1 machines around this office (even though it seems like ancient history). A few years from now I think its feasible that I'll be having similar memories of my Win2K machine as I post to DevShed from my Linux workstation.

This is a great thread and sorry to come is so late. But better late than never

I really like d-man initial comments about source-forge. Not only have a lot of the projects in source forge not gone so well, but source forge itself isn't doing to well. Check out this interview with Tim Purdue where he states Source Forge earns about as much a a gas station THere are also links to the new gforge project which looks a lot better with built in project management capabilities.

IMHO it isn't just the source forge projects that fail, it is a large percentage of all projects that fail The reasons that softare projects fail are the same: lack of planning, lack of proper documentation (specification, design docs, use me docs), lack of proper organisation (project management, engagement), lack of testing and lack of engagement with the client. Just as there are Open Source projects that have failed there are many that have succeeded: PHP, PostgreSQL, MySQL, Linux, BSD, PHPBB, etc...

Source Forge is a great idea, but I think that some further resources are needed like project management, but also documentation on how to run a project and how to write a specification. I think gforge goes a way to solving this, but not all the way. I would like to see a system up like this, with a healthy community supporting it.

I agree with rycamor that Open Source needs to stay "Open" and not try to standardise too heavily. I also think that as a whole we need to all ensure our applications can talk to each other. And this is where I harp on about XML as an API definition language to make our life easy.

<rant target=M$ solution=OOo/>


Now just read this M$ article. They still haven't got it, if you don't publish the damn API how can I write something that integrates with your product? Not only that but I wonder how good an API M$ products actually have since OOo (Open Office) wrote their M$ integration by trial-and-error -yuck. That is why I'm asking to OOo guys about using OOo server-side and the response:



Not bad, better than a kick in the nuts (which is better that writing that myself) as long as I have some cool php application server to manage resources. OOo actually integrates better with M$ than the rest of M$ office. LL

Here is my favourite quote from the article:


Sorry but you just don't understand man, I guess you haven't opened up a propreitary crappy fat Excel file in OOo Math and saved it as a master file. Perfect MathML including the equations. I can take this format and integrate it back into my XML-based Document Management System. They just don't understand that, and despite M$ office supporting XML internally, what does it matter if I can't integrate this with other systems I write. Surely they can't believe that all software is going to be .NET based so it doesn't matter? Or do you think this is their plan

M$ are also still in court with Sun:
http://news.com.com/2100-1001-269266.html
Because of their tidy sabotage work on JVM (Java Virtual Machine). And check out this terrible publicity for J2EE:

http://www.objectwatch.com/issue_42.htm

Not good there. Sure like to test this versus freebsd/apache/php (+srm?). So if java were to fall there would only be an Open Source alternative.

</rant topic="back on">

This makes our efforts even more important and believe that we desperately need a full Enterprise Open Source solution. If you look all the pieces are there:

*Open LDAP for your LAN User Management
*OPen Office
*Webdav as Mod_Dav an apache module for LAN access to file server (and integration)
* OPen Exchange server (not quite open) from SUSE (thanks Rycamor/Jpenn)
* FreeBSD as your server, combined with Apache, Postresgre, PHP and probably VL-SRM.

*XML related technologies including:

*XML Schema - define your docs
*RDF (Dublic Core) desribe them
*XSLT - translate doc between systems, or to different
*NCX - define the navigation structure of files
*and so on ....

We need a good outlook/exchange client program, and to agree on some api protocols, but beyond this the pieces are there. What are we waiting for?

My only question is this: what about licensing? I'm a small company operating in little NZ. It is very hard to get work I we have to compete against the big players. My onyl advantage is my toolbox If I give my docs (specification/design) and code out, I lose the little advantage I have. I can't afford to operate with licenses where everything is free otherwise we go under and I have to go back and work for the man Is there some middle ground here, where I can release some parts of my software as open source, and others not, as well as charging corporate users? Can you'all convince me to go all open source? Is there a good business model out there that works for OS products? Speak to me

I *can* think of these advantages:

* real reason to switch for companies
* number of users = testers
* when was the last time you opened up the php source code?
* recover costs with configuration , security, speed, support, donations, custom components, corporate sponsorship, hosting, training
* can use other os technologies and components, without dilemma, can take advantage of rest of community
* fast distribution
* fun, honest, ethical
* better testing and review
* bug reports, extensions

But still giving out the source scares me as does actually giving a core product away for free that I *need* to earn from. It is okay for Tom Lane (Postgres) and the OOo guys they get paid by Sun.

We're working on a Learning Management System (with XML-based DMS behind it) to compete with Blackboard locally, and we would have no problems if this system were to be used by the Open Source community to create courses for Open Source Developers (eg XUL). Hosting is fine here, but if we give out the source doesn't that make it easy for BB to catch up with us? On the other hand if everything we build sits on Open Source software (apache/php/postgres) and we even integrate Open Source products (phpbb) what implications does this have?

Interested to hear from you'all have to say about this.

Btw, all the best for Xmas and New Year to the devshed community

Z.

Hey Zitan,

I've read most of Roger Sessions' ObjectWatch newsletters and for those of you who haven't he's a big Microsoft supporter. He likes their development languages, their servers and the hardware they run on. Anything that Java can do, a Microsoft product can do better, cheaper and faster. This opinion precedes the .NET days. Of course they used the Pet Store which was not designed as a benchmark, but was designed to show all the different things Java could do. There have been several articles written on this before where it has been noted that the Java Pet Store is over-engineered and heavily abstracted since it wasn't intended to be a benchmark. The .NET Pet Store is just a bunch of thin wrappers over stored procedures. Its an apple-to-oranges comparison, but Microsoft still likes rolling it out and beating the Linux crowd over the head with the "proven fact" that their servers and code perform faster.

I would liked to see a comparison between .NET and the open source JBoss application server since it was the JavaWorld Editor's #1 choice over BEA and IBM Websphere. If you go to JBoss's website they have some interesting statistics on some of their customers. For example BASF chose JBoss, Apache, OpenLDAP, Postgresql, Red Hat and Tomcat running on both 32-bit and 64-bit Itanium servers for a shopping platform. Its designed to handle 100,000 subscribers with 25,000 concurrent users and still maintain 3-5 second response times. Microsoft makes some good products but I think open source can still compete with them.

You mentioned OpenExchange Server from SuSE. If you've never looked at it before, Bynari is another interesting company that while not free, is still a lot cheaper than Exchange.

My only question is this: what about licensing?
I favor licensing that allows you to create closed commercial applications if you want. If someone wants to pay me to write free code to give to others then I'm all for it, but as long as the bills keep coming, I have to make money somehow. One of the last open source programs I downloaded was the Firebird database. Its license said that you could distribute the compiled database binaries as part of a closed-source commercial application. If you changed the Firebird source code (one would assume to improve it), you had to make this improvement available to the Firebird development team. In other words, the base product was open, but custom applications could be built around it.

To me, this is a win-win-win situation. The developer gets to make a living, the customer gets a good product at a lower price (because you don't have to charge them licenses for the database) and the open source team gets improvements and bug reports. I don't see how I can make a living giving away all my source code. I would prefer to just try to provide customers with better value for their money: a better quality product at a much lower price than is currently available. One of my favorite products in UltraEdit . A huge amount of functionality for ~$30 produced by one guy (after five years he's had to add two more employees). This is the kind of product I want to produce.

PS - Merry Christmas to you to...

One quick note about my last post: It may confuse some readers that I started this thread talking about open source and wind up talking about making money on closed source applications. As a developer and open source user I think I have a responsibility to give back something to the community that has helped me a lot. I don't see it as a black-and-white decision where I have to be 100% open or 100% closed.

To expand on Zitan's question about licensing, I think how a product is licensed should be based on what the product is. I don't think there's a one-size-fits-all license for everything. You have to choose a license that is best / fair for everyone involved.

I like sharing but I do need money to get by. I don't need billions like Gates/Ellison but I do need something. Parts of something I develop may be 100% open. Other parts that are based on it and make my unique product may not.

Since I haven't got all the answers, this opinion is subject to change if someone can convince me otherwise.

I was reading Linux Magazine last night and came across a small newsbrief that Mitch Kapor, the guy who launched Lotus 1-2-3 and Lotus Notes, has started a new company called The Open Source Application Foundation(OSAF). He did this using $5 million of his own money to "create and gain wide adoption of Open Source application software of uncomprimising quality". My kind of guy!

Their first product is going to be a PIM written in Python/Tkinter code-name Chandler. They have a prototype called Vista that was written by Andy Hertzfeld a leader of the original Apple Macintosh development team. Currently their preferred toolkit includes Python, Mozilla, Jabber, Zope Object Database and wxWindows/wxPython. Their target platorms are Windows, Linux and Apple.

The website doesn't have a lot of pages and can be viewed in a few minutes. Definitely worth a look.