I don't think the writer took my words out of context or added too much spin - I think he basically got it. My only regret was that the term "Average Joe" was probably read by everyone well beyond the context we spoke in. I can't remember if I used it first or if the reporter did, but we weren't talking about the people who can't program a VCR. I would include in that category, though, Joes and Janes who in 1994 were using Mosiac's "view source" function to learn HTML and write their own pages. Or Joes and Janes who create animations in Flash. Or Joes and Janes who are creating more than mere wiki pages on WikiCalc or JotSpot.
Most of these people have never heard of compilers or CMM levels (bless 'em!). But they are programmers. I'll hesitate to call them engineers, only because that word should be saved for people with both formal training and certification; though I would wager that most people reading this comment with "engineer" in their employment title have neither of those.
The point I made to Phil, and that I think came out in the article, is that these Average Joe "programmers" are starting to work with each other, and with the "expert professionals", to create some very interesting things, both publicly and privately. If you're against that notion - that programming is a monastic science that requires years of experience before you're allowed to subject others to your work - then you're not only against the concept of Open Source, you're against what made the web successful ("view source", anyone can write HTML).
The license in the "open source" grant you only the right to use it for non commercial purpose.
Sorry, that would not be an Open Source license. The point of having software be Open Source is that it provides all the necessary components to being able to outlive its original authors. Such a license would prevent that.
A million dollar software project can be broken into a series of different projects, each just large enough that it can be spec'd out with a reasonable set of milestones (almost always less than 15-20). Software development schedules always change based on inherent risk, so you probably shouldn't spec out a 500 milestone project. And the Sponsor always has to be someone who can check the Developer's work - so this isn't exactly "put $15K in, get good solid code out overnight" kind of thing.
I would think that if you were the project manager for a 15 person software development project, you could probably parallelize 10-15 sub-projects in sourceXchange (or less if you have multiple developers per project), all working on different things. Of course it's not always nice and parallelizable, sometimes there's only work for a few at a time - but the beauty of this model is that you don't have to pay for 15 full time developers during those periods.
There's another thing a developer can do - be a Sponsor on a project where the person actually paying is upstream, and only knows their need, not how to define a solution. In other words, let's say you (average Slashdot reader) know of someone who has a need for some sort of development (say a Linux driver for the CCD camera in the Sony Vaio Picturebook) and is willing to pay to have it developed, yet doesn't know enough about Linux kernel internals to write a spec or approve milestones. You would do it, if fact you know how you'd do it, but you don't have nearly enough time. This model lets you be the "sponsor", acting as a subcontractor to this other someone who needs the driver written, who specifies (generally) what to build and how, and then outsources that actual development through sourceXchange, passing along the cost (plus an amount for your own time and effort) to the original customer.
This is a bit overdue, actually - and I'm confident that Eric isn't the first personality from this community who's been invited to speak to MS, though perhaps the first to have done so so publicly.
Microsoft's not dumb, and they're not afraid to change course, even drastically. If Win2K is a more or less total disaster, if Linux continues on its trajectory, and if there's a biz model behind it, I could see MS doing their own Linux distribution, porting MSOffice to it (though probably *not* open sourcing MSOffice), etc. Actually, all they need to do is port Win32 over to Linux, and they have a platform "story" to tell their developers.
Actually all Microsoft needs to do is buy time to build up their services division, as IBM has done over the last few years. Right now MS is way way overreliant on the high margins associated with proprietary software development - a model that more or less can not survive in the OSS world. Services can, though; but margins on services are much lower, so MS needs to figure out a services strategy that still retains a high margin. That's not easy to do, since you can't really protect yourself against your competitors offering the same service at a lower fee (and margin).
I appreciate all the comments that have been posted, particularly those from Frank Warmerdam and _Dodger. BTW, I'm the Brian involved with sXc, under my posting identity on/..
Most of the points I was going to make have been made; particularly the point about the fact that we are not a bounty system, but a project-based system, which means that the developer and sponsor agree to work together at the beginning, rather than the end, of the development cycle, and that the developer is paid according to a series of milestones in the project. This is the only sane way to do larger projects, and I think it addresses some of the original essay's author's comments.
Also, since announcing sXc, the single most often asked question, one way or another, is, "can developers (or anyone else) propose projects that can get funded". As time has gone on, we've realized this could actually be a much more effective way of generating funded RFP's - rather than trying to guess which companies would have projects they'd want to fund, we'd look at projects that the developers have identified as crucial, and try and figure out who we could convince to fund them. In our old thinking, this was the tail wagging the dog, but we now see this as really the dog wagging its tail.
And of course, long-term, the real funds for projects will come from the end-users who aggregate their funds for projects. Think about this - a conservative estimate on the size of the Linux user base is 12 million users. Imagine if each of them were to contribute $40 towards the evolution & development of Linux-related software (the kernel, utilities, graphical apps, etc) over the course of a year. That would be $480 million dollars available for projects!
Finally... keep in mind that there is tremendous flexibility for what can constitute a "milestone" in a project, and what a project itself can fund. It can be source code, but it can also be documentation, test suites, UI/installation, quality assurance... all those things that tranditionally haven't been done by developers because, frankly, they're usually pretty boring to do. We can also have projects for which each milestone is simply a marker in time: for example, let's say the project is "to monitor and maintain the 3C509 network card driver family in *BSD over the next 6 months". Let's say each milestone was simply that every two weeks, developer would report to sponsor what they did to fulfill that directive. Maybe there was a kernel change that mandated a patch; maybe there was a new model of a card released; maybe there was a new bug posted that needed addressing. This is a much looser kind of project, but it could be exactly what is needed to help make sure that card family doesn't suffer from bitrot long-term.
At any rate, this is definitely a new space that CoSource, sXc, and other efforts are entering, and we're going to make mistakes no doubt - but every day I get more optimistic that this is actually going to work. =)
Disclaimer: I'm one of the newer members of the OSI Board. I don't speak for them, at least right now I'm not.
After the APSL 1.0 situation, OSI started a mailing list to review licenses submitted to us for validation as Open Source. To subscribe, send mail to license-discuss-subscribe@opensource.org.
We had the chance to put OSI certification on the APSL 1.1 for their press release today; but because we weren't allowed to share the new license with license-discuss beforehand, we chose not to give it certification, even though we on the board were confident the new license would pass. But in the spirit of "all bugs are shallow given sufficient eyeballs", we thought it'd be best to see what the community thought of the proposed changes, specifically along the lines of conformance to the OSD.
Thanks,
Brian
Using this to build parallel computers...
on
Mini Board PC
·
· Score: 1
So, what I'd really like to do is buy something like this (something that can be the guts of a PC in the size of a 3.5" slot), buy maybe 10 of them and put them into a RAID rack-mount enclosure, and then plug all 10 into a set of SCSI disks and have myself a kick-ass parallel web server.
What I'm wondering is to what degree these units could share devices, for example, could I have a single PCI chain that all these units had ID's on? Of course, I could have just one unit act as the "file server" and have the others mount the disks through it via NFS or something, but that's a lot of overhead, if instead all the units could treat the disk as something "local", and get the speed advantage of that.
I'm thinking something like sharing a PCI chain is probably a hell of a lot easier to support than sharing RAM. But that's OK - I'm happy dealing with this administratively like 10 separate machines, because I'm sure I could write scripts to manage that (e.g. restart the web servers on all 10 simultaneously). All I want is something that looks and acts like 10 rack-mount servers in the space of 1 - if we it can share a PCI chain, so much the better.
It would make sense for them to do this at the Coronet in SF. That's where they debuted Star Wars in 1977 and the re-release in 1997. That's where I'm hoping to see it, early release or not, because of the great seats and great sound system.
Don't get your panties in a bunch. Performance numbers mean very little, except that under some theoretical load (and they at least superficially attempted to equalize variables between the systems) the system should be able to perform "well enough" to be considered a server OS, on par with other servers below $10K. That's all they needed to prove, but of course for the press to report it it has to be "the fastest".
A fair test is to start with these roughly comparable systems, as Apple did, and then give each out-of-the-box fresh system to a team of platform experts (e.g., give the Sun box to two Sun employees, the NT box to MS employees, the Linux box to Dean Gaudet, etc:) and give them 2 hours to configure and optimize the environment. Then run the benchmarks. I bet each will perform about 10 times higher than the numbers found for each in Apple's tests.
My point is that sometimes the choice of platform for a project should factor in the expertise in a group - if your co-workers are Solaris weenies, well by all means go with Solaris.
The real debate should be, given out-of-the-box configurations, which platform & server are easiest to speed-tune for the non-expert? Of course, that's a hard number to give 5 degrees of precision to.
I don't follow your logic. Allow "any and all TLD's" and the domain name speculation problem will just skyrocket. For example, I'll go register ".con" (pun intended) and start registering "mcdonalds.con", "yahoo.con", etc, since that's probably a common mispelling of ".com".
I tend to believe there's no good reason to create *any* new TLD's, outside of country-specific ones.
Slashdot Mirror
I don't think the writer took my words out of context or added too much spin - I think he basically got it. My only regret was that the term "Average Joe" was probably read by everyone well beyond the context we spoke in. I can't remember if I used it first or if the reporter did, but we weren't talking about the people who can't program a VCR. I would include in that category, though, Joes and Janes who in 1994 were using Mosiac's "view source" function to learn HTML and write their own pages. Or Joes and Janes who create animations in Flash. Or Joes and Janes who are creating more than mere wiki pages on WikiCalc or JotSpot.
Most of these people have never heard of compilers or CMM levels (bless 'em!). But they are programmers. I'll hesitate to call them engineers, only because that word should be saved for people with both formal training and certification; though I would wager that most people reading this comment with "engineer" in their employment title have neither of those.
The point I made to Phil, and that I think came out in the article, is that these Average Joe "programmers" are starting to work with each other, and with the "expert professionals", to create some very interesting things, both publicly and privately. If you're against that notion - that programming is a monastic science that requires years of experience before you're allowed to subject others to your work - then you're not only against the concept of Open Source, you're against what made the web successful ("view source", anyone can write HTML).
Brian
Brian
Sorry, that would not be an Open Source license. The point of having software be Open Source is that it provides all the necessary components to being able to outlive its original authors. Such a license would prevent that.
Brian
A million dollar software project can be broken into a series of different projects, each just large enough that it can be spec'd out with a reasonable set of milestones (almost always less than 15-20). Software development schedules always change based on inherent risk, so you probably shouldn't spec out a 500 milestone project. And the Sponsor always has to be someone who can check the Developer's work - so this isn't exactly "put $15K in, get good solid code out overnight" kind of thing.
I would think that if you were the project manager for a 15 person software development project, you could probably parallelize 10-15 sub-projects in sourceXchange (or less if you have multiple developers per project), all working on different things. Of course it's not always nice and parallelizable, sometimes there's only work for a few at a time - but the beauty of this model is that you don't have to pay for 15 full time developers during those periods.
There's another thing a developer can do - be a Sponsor on a project where the person actually paying is upstream, and only knows their need, not how to define a solution. In other words, let's say you (average Slashdot reader) know of someone who has a need for some sort of development (say a Linux driver for the CCD camera in the Sony Vaio Picturebook) and is willing to pay to have it developed, yet doesn't know enough about Linux kernel internals to write a spec or approve milestones. You would do it, if fact you know how you'd do it, but you don't have nearly enough time. This model lets you be the "sponsor", acting as a subcontractor to this other someone who needs the driver written, who specifies (generally) what to build and how, and then outsources that actual development through sourceXchange, passing along the cost (plus an amount for your own time and effort) to the original customer.
There's lots of flexibility in the system.
Thanks.
Brian
This is a bit overdue, actually - and I'm confident that Eric isn't the first personality from this community who's been invited to speak to MS, though perhaps the first to have done so so publicly.
Microsoft's not dumb, and they're not afraid to change course, even drastically. If Win2K is a more or less total disaster, if Linux continues on its trajectory, and if there's a biz model behind it, I could see MS doing their own Linux distribution, porting MSOffice to it (though probably *not* open sourcing MSOffice), etc. Actually, all they need to do is port Win32 over to Linux, and they have a platform "story" to tell their developers.
Actually all Microsoft needs to do is buy time to build up their services division, as IBM has done over the last few years. Right now MS is way way overreliant on the high margins associated with proprietary software development - a model that more or less can not survive in the OSS world. Services can, though; but margins on services are much lower, so MS needs to figure out a services strategy that still retains a high margin. That's not easy to do, since you can't really protect yourself against your competitors offering the same service at a lower fee (and margin).
We live in interesting times.
Brian
I appreciate all the comments that have been posted, particularly those from Frank Warmerdam and _Dodger. BTW, I'm the Brian involved with sXc, under my posting identity on /..
Most of the points I was going to make have been made; particularly the point about the fact that we are not a bounty system, but a project-based system, which means that the developer and sponsor agree to work together at the beginning, rather than the end, of the development cycle, and that the developer is paid according to a series of milestones in the project. This is the only sane way to do larger projects, and I think it addresses some of the original essay's author's comments.
Also, since announcing sXc, the single most often asked question, one way or another, is, "can developers (or anyone else) propose projects that can get funded". As time has gone on, we've realized this could actually be a much more effective way of generating funded RFP's - rather than trying to guess which companies would have projects they'd want to fund, we'd look at projects that the developers have identified as crucial, and try and figure out who we could convince to fund them. In our old thinking, this was the tail wagging the dog, but we now see this as really the dog wagging its tail.
And of course, long-term, the real funds for projects will come from the end-users who aggregate their funds for projects. Think about this - a conservative estimate on the size of the Linux user base is 12 million users. Imagine if each of them were to contribute $40 towards the evolution & development of Linux-related software (the kernel, utilities, graphical apps, etc) over the course of a year. That would be $480 million dollars available for projects!
Finally... keep in mind that there is tremendous flexibility for what can constitute a "milestone" in a project, and what a project itself can fund. It can be source code, but it can also be documentation, test suites, UI/installation, quality assurance... all those things that tranditionally haven't been done by developers because, frankly, they're usually pretty boring to do. We can also have projects for which each milestone is simply a marker in time: for example, let's say the project is "to monitor and maintain the 3C509 network card driver family in *BSD over the next 6 months". Let's say each milestone was simply that every two weeks, developer would report to sponsor what they did to fulfill that directive. Maybe there was a kernel change that mandated a patch; maybe there was a new model of a card released; maybe there was a new bug posted that needed addressing. This is a much looser kind of project, but it could be exactly what is needed to help make sure that card family doesn't suffer from bitrot long-term.
At any rate, this is definitely a new space that CoSource, sXc, and other efforts are entering, and we're going to make mistakes no doubt - but every day I get more optimistic that this is actually going to work. =)
Thanks for the comments, all.
Brian
p.s. - the first draft of the sXc developer agreement is now online. Comments welcome..
....not to give interviews before 9am in the morning! =)
Brian
Disclaimer: I'm one of the newer members of the OSI Board. I don't speak for them, at least right now I'm not.
After the APSL 1.0 situation, OSI started a mailing list to review licenses submitted to us for validation as Open Source. To subscribe, send mail to license-discuss-subscribe@opensource.org.
We had the chance to put OSI certification on the APSL 1.1 for their press release today; but because we weren't allowed to share the new license with license-discuss beforehand, we chose not to give it certification, even though we on the board were confident the new license would pass. But in the spirit of "all bugs are shallow given sufficient eyeballs", we thought it'd be best to see what the community thought of the proposed changes, specifically along the lines of conformance to the OSD.
Thanks,
Brian
So, what I'd really like to do is buy something like this (something that can be the guts of a PC in the size of a 3.5" slot), buy maybe 10 of them and put them into a RAID rack-mount enclosure, and then plug all 10 into a set of SCSI disks and have myself a kick-ass parallel web server.
:)
What I'm wondering is to what degree these units could share devices, for example, could I have a single PCI chain that all these units had ID's on? Of course, I could have just one unit act as the "file server" and have the others mount the disks through it via NFS or something, but that's a lot of overhead, if instead all the units could treat the disk as something "local", and get the speed advantage of that.
I'm thinking something like sharing a PCI chain is probably a hell of a lot easier to support than sharing RAM. But that's OK - I'm happy dealing with this administratively like 10 separate machines, because I'm sure I could write scripts to manage that (e.g. restart the web servers on all 10 simultaneously). All I want is something that looks and acts like 10 rack-mount servers in the space of 1 - if we it can share a PCI chain, so much the better.
Anyone remember Sequent?
Brian
It would make sense for them to do this at the Coronet in SF. That's where they debuted Star Wars in 1977 and the re-release in 1997. That's where I'm hoping to see it, early release or not, because of the great seats and great sound system.
Brian
Don't get your panties in a bunch. Performance
:) and give them 2 hours to configure and optimize the environment. Then run the benchmarks. I bet each will perform about 10 times higher than the numbers found for each in Apple's tests.
numbers mean very little, except that under some
theoretical load (and they at least superficially
attempted to equalize variables between the systems) the system should be able to perform
"well enough" to be considered a server OS, on par with other servers below $10K. That's all they needed to prove, but of course for the press to report it it has to be "the fastest".
A fair test is to start with these roughly comparable systems, as Apple did, and then give each out-of-the-box fresh system to a team of platform experts (e.g., give the Sun box to two Sun employees, the NT box to MS employees, the Linux box to Dean Gaudet, etc
My point is that sometimes the choice of platform for a project should factor in the expertise in a group - if your co-workers are Solaris weenies, well by all means go with Solaris.
The real debate should be, given out-of-the-box configurations, which platform & server are easiest to speed-tune for the non-expert? Of course, that's a hard number to give 5 degrees of precision to.
Brian
I don't follow your logic. Allow "any and all TLD's" and the domain name speculation problem will just skyrocket. For example, I'll go register ".con" (pun intended) and start registering "mcdonalds.con", "yahoo.con", etc, since that's probably a common mispelling of ".com".
I tend to believe there's no good reason to create *any* new TLD's, outside of country-specific ones.