What they have is 1. a 15mph limit 2. give way signs that force incoming traffic to wait for the outbound stuff
One of the thing the article doesnt discuss is that the road is the best way to get from west bristol to the airport -and all us locals know that. The real problem is the airport itself is expanding at a rate that isnt sustainable.
AFAIK, the sole patents are -as you point out- on mechanisms for bytecode verification. But there is no guarantee that google dont have a different solution to the problem.
To make life more complex, Sun are actually breaking their own Java Community Process rules by refusing to give apache access to the test kit for Java, so that the Harmony team can test their clean room implementation of the java 6 classes. Passing that TCK automatically grants patent rights, so by denying access to the test kit, sun are stopping Harmony declaring that it is an implementation of Java. (Note that Sun do offer access to the TCK for GPL-licensed open source Java runtimes that are mostly derived from Sun's own code; they dont want any other open source implementation to be called Java).
So: sun broke the contract with apache, but apache lack the lawyers to do anything about it. Now Sun may end up going up against intel and google. They'd be better of giving up the TCK to the harmony team, so that they can at least retain some control over what ships.
-steve
(apache member, but not involved in android, harmony or similar)
There was an early-day motion by some 50+ MPs saying "extend it for the sake of the artists". My own mp was one, so I complained and got an apology back -once he understood the details more he had removed is name from the motion.
I guess the bribes werent big enough, not when you are an MP in a university town with all those students to upset
1. its a behind the firewall toy; limited offline support, no good for cross-company collaboration
2. It doesnt like firefox being the default browser
3. Its windows only
4. It has really good integration inside word and excel; you can see who else is online, etc.
5. The install process is complex; the only thing worse than it is probably MS project server, which installs atop sharepoint.
Its half-way towards real-time collaboration on document editing, but you can still only have one person editing simultaneously (in my limited experience), and it is too windows based.
If the collaboration features in Office were designed to be back end neutral, then it would be interesting -you could plug in something better behind word. Now that google docs does low-grade editing through firewalls, you dont need anything on the client at all, though the doc quality is pretty low.
My favourite collab tool: Confluence. Wikis done right.
Indifferent to it and all the other OSS surveys you get, normally by some student mining the big OSS SCM repositories and thinking they will be the first person to survey all the developers to find out how they work. Why, its almost as bad as the in-house cross-organisation architecture mailing lists that you end up on if you do closed-source.
The authors mailed 300+ people and only 34 replied. From the responses of the people they decided whether they were FSF believers, pragmatists (e.g. BSD people) and inferred the opinions of GPLv3 from that.
I think their conclusion that BSD/apache people won't suddenly embrace GPLv3 is probably valid, but you don't need to do a survey for that. And a survey can't determine which is better, GPL versus BSD, because its such a religious issue.
I write technical books; my last one is obsolete because it is 5 years old. Clearly I dont give a fuck about extending copyright after my death.
But literar authors seem different, maybe because they think their work is somehow more valuable. The NYT writer even had the audacity to cite Sylvia Plath, who so cruelly brought the 70 years after death deadline forward by killing herself, so depriving her offspring of their rightful revenue.
whenever its music copyright, they bring up the musicians and their widows, because talking about depriving record companies of a big fraction of their cocaine budget doesnt build up sympathy.
Java, for all its strengths server-side, is still hard to write for on the client. Swing, SWT, AWT: all are hard to write and test. Applets suck, and Java Web Start, although it works (even on linux), isnt that common.
If you look at the popularity of Ajax sites its because
-it works on all three platforms: IE, Firefox and Safari
-it lets you extend existing sites instead of moving to a new platform
-there's no download/startup hit
-it lets you had off to Flash the multimedia stuff
Silverlight is effectively MS promising to port a subset of Avalon to the Mac. Avalon is an XML format for the GUI, backed by.NET code, or remote code. Its goal is to let you write rich apps, more easily than today.
For OSS, we have something equivalent: XUL. There is good JavaScript+XUL support, and some Java+XUL integration, but it doesnt have the full backing of people like Sun (yet).
You can still host the old software on your new box, just by running it under VMWAre or Xen and configuring the VM to have less CPUs than your real box. You may need a bit more RAM, but RAM costs less than PDFLIb upgrades.
The apple (and presumably) BSD version of snprintf() always zeros the last byte. But even zeroing the end of the string doesnt avoid problems, because information has been dropped -unless the caller checks for it they will carry on, not knowing that half their string has just been discarded. IT may be robust against buffer overflows, but it has a datacorruption bug that can take a while to surface. And there is nothing like a large database with invalid data to cause nightmares.
But yes, I do take your point about rigorous coding. That is not something Microsoft were known for in the past. As a word doc can include any activeX control or other embeddable OLE object, objects that load their state from binary data serialized to the file, there's a whole truckload of buffer overflows out there, just waiting to be found.
From my past days doing C++ code, I recall that some snprintf() implementations don't zero the last byte in the buffer if they overflow...there's been no overwrite, but the string is toast. So its not enough to call snprintf() and get overconfident that your app is secure...you need to check for an overflow and bail out if there is an error, or go back and allocate more.
Code that just does snprintf() and doesnt check the results is not vulnerable to buffer overflow attacks, but you can damage a string which can be used in other functions, so lead to trailing damage.
The safe way to do string concatenation is through half-decent string classes, ones that allocate on the fly. cstring, perhaps.
In the 'stricter' languages: Java, C#, python, lisp, you dont have buffers, you dont have buffer or array overflow. Life is simpler. Instead you get some kind of runtime fault when an out of bounds event happens. I think this is what MS have been trying to retrofit into their C++ code. The main problem here is that when C++ faults with something like a pointer error, your entire memory footprint is potentially contaminated. Whereas when a Java program goes wrong, its damage is limited to all the stuff in scope (and the things indexed off it) at that point in time. This is why nobody bothers to go for buffer overflows when attacking a Java app -instead you assume the coders forgot to escape web page parameters being used in SQL statements and go straight for SQL injection attacks. The language has moved the vulnerabilities.
One diff between UK and US navigation is that US residents know the street names of all roads they use. To aid this they are often named after presidents or (in portland, or), Simpsons characters.
In the UK ten+ streets in the same area can all take the same first name and vary only on the ending (as you noticed), or be completely randomly named. To stop people learning the names the signs are hidden instead of placed somewhere obvious and illuminated.
In the UK this means that (1) nobody knows where the road you are looking for is, even if they are standing in it, and (b) directions tend to be based on landscape features (shops, pubs, junctions) rather than "5-th and harrison" or "monroe and 23". Unfortunately, people's memories are unreliable. When they give "take the third left then the second right" they forget about that minor left turn or the other fork in the road, so the directions are worthless.
As a consequence, you cannot trust anyone you ask to know where you want to go, or, if they do, to give you directions you can use.
If you look at the RIAA's tactics, its an attempt to remain relevant in a world where their core value "ability to control the supply chain to the retail outlets" is gone.
1. Attempts (successful in the US, not yet in the UK) to extend copyright. ==retain revenue from legacy products.
2. Attempts (like this) to maximise revenue from new content sales.
3. Attempts to impose a tax on all media-enabled devices, a tax which doesnt correlate to any track sales, so is probably exempt from the need to give the songwriters, publishers or artists any money.
#3 is the most insidious, because despite the tax, there's no guarantee you get anything for it. After all, in the UK we pay a tax on blank cassettes ("home taping is killing music"), yet they still dont like you copying it.
If they can get a tax out of every MP3 player, then they will go for the DVD-RW drives and the PCs, and then the broadband connection. While the publishers will go after the printers...
Oh, that's so unfair. Go look at this little bunny I wrote last year, pointing out the entire Web Service SOAP stack and its belief in seamless mapping between Java/C# and XML was a load of fundamentally unachievable bollocks.
When I was at the IEEE conf presenting it (and getting best paper, BTW), I had to put up with three days of academics stuck in the depths of their little web service, none of whom seemed to step back and notice that what they basing their work on was junk. Instead they were using Apache Axis or similar and repeating exactly the same mistakes enterprise developers do: they believed IBM and Microsoft knew what was best for them.
I actually prefer open source conferences. Good talks, good audience, ubiquitous beer.
I work in corporate R&D, so want to make one ruthless statement to a pretty impressive list of publications.
In academia, publications are your metric of success. In industry, its your ability to generate an ROI by improving the company's profits.
Xerox PARC was a failure to xerox. Not to Canon, HP and adobe (laser printing and PDF), not to apple, MS or the rest of the world that uses GUIs. But it was to Xerox. I dont know if MSR has repaid their investment yet. Frankly, at the estimated $10B that vista cost in full time engineering staff, I dont know when it will repay its investment either. At least MSR did some nice papers for other people to read and make use of.
One issue with hiring the masters like Lamport is that they like to do their own thing, and want a large staff underneath; in industry its harder to get head count than in academia, where you have undergrads, RAs and phd students to suffer at low cost for the sake of a professor.
The other problem -tech transfer- is the enemy of all R&D labs, and of academia too. There's a lot of good ideas out there, that don't make it out into a world that has the x86 as the primary CPU, A DOS derivative and a Unix derivative as the choices of OS, and C/C++ as the primary programming languages.
FWIW, I work in a corporate R&D lab in the UK, and getting anything taken up is always a miracle to be celebrated. Except when it takes so long to come to market that they shouldnt have bothered. This is why open source is so much better as a way of doing tech transfer. If you have something good, a patch, a test and the ability to argue your case, it can be in the code tree in a week, and in people's hands the next day, in mainstream distros within a month or two.
Whereas MS Research? Vista took 5 years. Every new idea in the last three of those years will have been postponed to its successor. So the lag between an idea and product is 3-5 years, compared to 3-5 weeks.
I actually think for consumers, a media centre PC with Vista Ultimate is a very nice toy. good gaming, good media experience within the DRM sandbox they have constructed (one that seemingly excludes Zune, which can only be good). For that reason, its a real screwup to have missed christmas, which would have been the key time to launch it at the masses.
For enterprise, there are some bits that appeal. Bitlocker could be good for laptop security, though TPMs can be used to partially encrypt bits of an NTFS drive today. (they can keep the key, but all the file metadata is public, along with the swap and hibernate files). The new power management modes may bring the OS up to par with MacOS X, which again is good for laptops, as will be bluetooth support (finally!) and better WLAN. all of these are laptop features, which is important. In the F500 business world, the laptop is the primary machine. It costs the same, its support costs are higher, but the extra 2-5 hours of employee work you get per week soon pays for it.
All the corporates really need is a version of XP with disk security, decent bluetooth and power management, and good mobile networking, plus office apps that dont handle going off the LAN so badly, and a web browser that is not so prone to buffer overflows. In my F100 day job, firefox has a really bad experience, because so many must-use websites were written about 8 years ago and have hard coded assumptions about IE and its scripting language built in. Opera works much better. And you know what? IE7 has a hard time too. To the extend that not only is the automatic update disabled, if the PC audit tools find you have installed it voluntarily, you get email telling you off. They are going to rework and retest their sites for IE7, and only then can we upgrade to IE7 or vista, which wont be for seven months minumum.
Its that cost of upgrading every web site, as well as every corporate app, that is going to make adopting vista expensive, even if the system management costs are less.
And Me? I'm really looking forward to the upgraded web sites, as they should be firefox compatible too. Then I can stop running WinXP in a VMWare image under ubuntu.
When I worked at a consumer PC vendor, along with the normal hard to fix incidents "sound goes off on mechwarrior on level 7", which take a while to replicate, those early adult DVDs caused a lot of problems on the first DVD-enabled PCS. These where the generation-1 DVD players, trying to render everything on a Pentium II/233 or maybe even a 266 MHz unit. Those adult DVD with their multi viewpoint features, and the need of the audience to freeze frames, caused a lot of problems. After about six months there was a shelf full of adult disks that were hard to play, alongside mainstream content that had high sustained bandwith, funny audio channels, complex menus etc. Funnily enough, the engineers were always willing to test the adult disks on new hardware and device drivers; it was the most thorough support team I ever knew. It was like they had a deep fear of regression failures creeping in. Word 97 not printing to postscript printers, AOL dialup failing, these were low priority compared to the different viewpoints in the orgy scenes working.
As someone who works on Apache ant, yes, we like bugreps that are replicable, and prefer patches with tests.
At the same time, we try and test our stuff, internally and externally. But the moment an x.0 release ships, we still get lots of bugreps. And you know why that is? Because when the x.0 release ships, a lot more people grab the app and use it. And unlike beta testers, these are not experienced developers. They are people who (in the Java context) dont know that the CLASSPATH env variable is a recipie for disaster, that you shouldnt have trailing backslashes or inner quotes in it. We have people whose Windows PC is an inconsistent mess and things just dont work on them. We get people who are running jpackaged and self-installed ant distros side by side, and get surprised that ant.sh delegates to jpackage installations, so the upgrade doesnt appear to take.
The issue is not that we dont beta test our software, it is that the testers, having a certain level of competence/experience, don't set up the apps in a pathologically bad way. Its not that the code doesn't work, it is that we cannot test all configurations, and that is what burns us.
Operating systems have the same problem only multiplied.
One thing I dont agree with is closing bugs unless they are fixed, or unless the team has made a WONTFIX decision. The troublespot is WORKSFORME, because, yes, that is the problem: code that works on some configurations and not others. There is a great ongoing bugrep in Eclipse, that says "LATER bugs get ignored", which is how that team works. Marking something as later doesnt just postpone the fix, it hides it. In Ant, we leave all bugs open until closed propery. Which is why we have 500+ bugreps right now, I guess:(
That's a really good summary. RHEL are in the top 3 contributors to the linux codebase, next to linus and IBM. Nobody else understands it as well, or (outside Linus), has the ability to effect change. That's important, because it means redhat are one of the few groups who can set a direction for linux that suits them.
With the JBoss purchase, they even have influence in many java projects. Not all; I suspect google sun and IBM have more. but they own hibernate and jboss, and have a presence in tomcat.
I'm busy moving from Suse to ubuntu, then from jboss to Sun glassfish. not for idelogical reasons, but for technical ones, like "they work really well"
Windows pricing is hard to compare, because its so variable. A copy of XP pro over the three year life of a PC is less than RHWS; but add on the cost of CAL licenses to network infrastructure (exchange, file servers) and the price of a PC rises. Same for server installs -the cost of the database scales with the number of users you want to support.
On the other hand, if you are a developer with an MSDN subscription, you get all of windows, office, much of visual studio for a flat rate, less than an AS subscription. Microsoft treats developers well.
But Fedora? To increase the value of the paid products, RH had to reduce the value of the free ones. Fedora achieves that. I dont see why we 'community' developers, whose of us who write the stuff that shipps with RHAS, have to suffer on an unstable distro with no support for products like VMWare. Here I am, working on things for no money, and yet RHEL still treat me as a leech who deserves low stability junk. If I want that, I'd use MS Vista:)
In car players are still alive. the CD as a distribution format for playback in those systems still works. Only now you can burn a disk on demand, from your collection.
Somewhere up in my attic I have a pile of legacy analog audio technology. a record player, the LPs, a VCR and its tapes, a cassette recorder. The most valuable item is my grandfather's tube-based AM radio from 1938; I may bring that down and use it as a PC speaker. The rest is just lost space. Which is the problem I have with CDs. What do you do with them? Unless you want to show off your legacy use case, they just collect in a box?
yes. Every post-CD format (minidisk, DAT, SACD, DVD-Audio) has been crippled by DRM that reduced the value of the format. CD is the best format for content because they can sell it through the legacy retail channel (shops), and because it is mostly DRM free. Now, if a record company embraced DRM-free content, at a reasonable encoding and price, things would be different. After all, its not exactly hard to go from DRM protected content to unprotected, so why do they bother? A willingness to give iTunes total control of the business?
If every CD only has two good tracks on it, then the real value of those two tracks is $10 each. The other eight are giveaways. The record companies want variable pricing so they can charge more for those two tracks. The alternative strategy: try and sell more hasn't occurred to them.
Slashdot Mirror
What they have is
1. a 15mph limit
2. give way signs that force incoming traffic to wait for the outbound stuff
One of the thing the article doesnt discuss is that the road is the best way to get from west bristol to the airport -and all us locals know that. The real problem is the airport itself is expanding at a rate that isnt sustainable.
AFAIK, the sole patents are -as you point out- on mechanisms for bytecode verification. But there is no guarantee that google dont have a different solution to the problem.
To make life more complex, Sun are actually breaking their own Java Community Process rules by refusing to give apache access to the test kit for Java, so that the Harmony team can test their clean room implementation of the java 6 classes. Passing that TCK automatically grants patent rights, so by denying access to the test kit, sun are stopping Harmony declaring that it is an implementation of Java. (Note that Sun do offer access to the TCK for GPL-licensed open source Java runtimes that are mostly derived from Sun's own code; they dont want any other open source implementation to be called Java).
So: sun broke the contract with apache, but apache lack the lawyers to do anything about it. Now Sun may end up going up against intel and google. They'd be better of giving up the TCK to the harmony team, so that they can at least retain some control over what ships.
-steve
(apache member, but not involved in android, harmony or similar)
There was an early-day motion by some 50+ MPs saying "extend it for the sake of the artists". My own mp was one, so I complained and got an apology back -once he understood the details more he had removed is name from the motion.
I guess the bribes werent big enough, not when you are an MP in a university town with all those students to upset
I have to use Sharepoint at work sometimes.
1. its a behind the firewall toy; limited offline support, no good for cross-company collaboration
2. It doesnt like firefox being the default browser
3. Its windows only
4. It has really good integration inside word and excel; you can see who else is online, etc.
5. The install process is complex; the only thing worse than it is probably MS project server, which installs atop sharepoint.
Its half-way towards real-time collaboration on document editing, but you can still only have one person editing simultaneously (in my limited experience), and it is too windows based.
If the collaboration features in Office were designed to be back end neutral, then it would be interesting -you could plug in something better behind word. Now that google docs does low-grade editing through firewalls, you dont need anything on the client at all, though the doc quality is pretty low.
My favourite collab tool: Confluence. Wikis done right.
Indifferent to it and all the other OSS surveys you get, normally by some student mining the big OSS SCM repositories and thinking they will be the first person to survey all the developers to find out how they work. Why, its almost as bad as the in-house cross-organisation architecture mailing lists that you end up on if you do closed-source.
The authors mailed 300+ people and only 34 replied. From the responses of the people they decided whether they were FSF believers, pragmatists (e.g. BSD people) and inferred the opinions of GPLv3 from that.
I think their conclusion that BSD/apache people won't suddenly embrace GPLv3 is probably valid, but you don't need to do a survey for that. And a survey can't determine which is better, GPL versus BSD, because its such a religious issue.
I write technical books; my last one is obsolete because it is 5 years old. Clearly I dont give a fuck about extending copyright after my death.
But literar authors seem different, maybe because they think their work is somehow more valuable. The NYT writer even had the audacity to cite Sylvia Plath, who so cruelly brought the 70 years after death deadline forward by killing herself, so depriving her offspring of their rightful revenue.
whenever its music copyright, they bring up the musicians and their widows, because talking about depriving record companies of a big fraction of their cocaine budget doesnt build up sympathy.
-steve loughran,
author, Ant in Action.
Java, for all its strengths server-side, is still hard to write for on the client. Swing, SWT, AWT: all are hard to write and test. Applets suck, and Java Web Start, although it works (even on linux), isnt that common.
.NET code, or remote code. Its goal is to let you write rich apps, more easily than today.
If you look at the popularity of Ajax sites its because
-it works on all three platforms: IE, Firefox and Safari
-it lets you extend existing sites instead of moving to a new platform
-there's no download/startup hit
-it lets you had off to Flash the multimedia stuff
Silverlight is effectively MS promising to port a subset of Avalon to the Mac. Avalon is an XML format for the GUI, backed by
For OSS, we have something equivalent: XUL. There is good JavaScript+XUL support, and some Java+XUL integration, but it doesnt have the full backing of people like Sun (yet).
You can still host the old software on your new box, just by running it under VMWAre or Xen and configuring the VM to have less CPUs than your real box. You may need a bit more RAM, but RAM costs less than PDFLIb upgrades.
The apple (and presumably) BSD version of snprintf() always zeros the last byte. But even zeroing the end of the string doesnt avoid problems, because information has been dropped -unless the caller checks for it they will carry on, not knowing that half their string has just been discarded. IT may be robust against buffer overflows, but it has a datacorruption bug that can take a while to surface. And there is nothing like a large database with invalid data to cause nightmares.
But yes, I do take your point about rigorous coding. That is not something Microsoft were known for in the past. As a word doc can include any activeX control or other embeddable OLE object, objects that load their state from binary data serialized to the file, there's a whole truckload of buffer overflows out there, just waiting to be found.
From my past days doing C++ code, I recall that some snprintf() implementations don't zero the last byte in the buffer if they overflow...there's been no overwrite, but the string is toast. So its not enough to call snprintf() and get overconfident that your app is secure...you need to check for an overflow and bail out if there is an error, or go back and allocate more.
Code that just does snprintf() and doesnt check the results is not vulnerable to buffer overflow attacks, but you can damage a string which can be used in other functions, so lead to trailing damage.
The safe way to do string concatenation is through half-decent string classes, ones that allocate on the fly. cstring, perhaps.
In the 'stricter' languages: Java, C#, python, lisp, you dont have buffers, you dont have buffer or array overflow. Life is simpler. Instead you get some kind of runtime fault when an out of bounds event happens. I think this is what MS have been trying to retrofit into their C++ code. The main problem here is that when C++ faults with something like a pointer error, your entire memory footprint is potentially contaminated. Whereas when a Java program goes wrong, its damage is limited to all the stuff in scope (and the things indexed off it) at that point in time. This is why nobody bothers to go for buffer overflows when attacking a Java app -instead you assume the coders forgot to escape web page parameters being used in SQL statements and go straight for SQL injection attacks. The language has moved the vulnerabilities.
oh, london cab drivers know how to come up with any excuse not to go south of the river.
One diff between UK and US navigation is that US residents know the street names of all roads they use. To aid this they are often named after presidents or (in portland, or), Simpsons characters.
In the UK ten+ streets in the same area can all take the same first name and vary only on the ending (as you noticed), or be completely randomly named. To stop people learning the names the signs are hidden instead of placed somewhere obvious and illuminated.
In the UK this means that (1) nobody knows where the road you are looking for is, even if they are standing in it, and (b) directions tend to be based on landscape features (shops, pubs, junctions) rather than "5-th and harrison" or "monroe and 23". Unfortunately, people's memories are unreliable. When they give "take the third left then the second right" they forget about that minor left turn or the other fork in the road, so the directions are worthless.
As a consequence, you cannot trust anyone you ask to know where you want to go, or, if they do, to give you directions you can use.
-steve
If you look at the RIAA's tactics, its an attempt to remain relevant in a world where their core value "ability to control the supply chain to the retail outlets" is gone.
1. Attempts (successful in the US, not yet in the UK) to extend copyright. ==retain revenue from legacy products.
2. Attempts (like this) to maximise revenue from new content sales.
3. Attempts to impose a tax on all media-enabled devices, a tax which doesnt correlate to any track sales, so is probably exempt from the need to give the songwriters, publishers or artists any money.
#3 is the most insidious, because despite the tax, there's no guarantee you get anything for it. After all, in the UK we pay a tax on blank cassettes ("home taping is killing music"), yet they still dont like you copying it.
If they can get a tax out of every MP3 player, then they will go for the DVD-RW drives and the PCs, and then the broadband connection. While the publishers will go after the printers...
-steve
Oh, that's so unfair. Go look at this little bunny I wrote last year, pointing out the entire Web Service SOAP stack and its belief in seamless mapping between Java/C# and XML was a load of fundamentally unachievable bollocks.
When I was at the IEEE conf presenting it (and getting best paper, BTW), I had to put up with three days of academics stuck in the depths of their little web service, none of whom seemed to step back and notice that what they basing their work on was junk. Instead they were using Apache Axis or similar and repeating exactly the same mistakes enterprise developers do: they believed IBM and Microsoft knew what was best for them.
I actually prefer open source conferences. Good talks, good audience, ubiquitous beer.
I work in corporate R&D, so want to make one ruthless statement to a pretty impressive list of publications.
In academia, publications are your metric of success. In industry, its your ability to generate an ROI by improving the company's profits.
Xerox PARC was a failure to xerox. Not to Canon, HP and adobe (laser printing and PDF), not to apple, MS or the rest of the world that uses GUIs. But it was to Xerox. I dont know if MSR has repaid their investment yet. Frankly, at the estimated $10B that vista cost in full time engineering staff, I dont know when it will repay its investment either. At least MSR did some nice papers for other people to read and make use of.
One issue with hiring the masters like Lamport is that they like to do their own thing, and want a large staff underneath; in industry its harder to get head count than in academia, where you have undergrads, RAs and phd students to suffer at low cost for the sake of a professor.
The other problem -tech transfer- is the enemy of all R&D labs, and of academia too. There's a lot of good ideas out there, that don't make it out into a world that has the x86 as the primary CPU, A DOS derivative and a Unix derivative as the choices of OS, and C/C++ as the primary programming languages.
FWIW, I work in a corporate R&D lab in the UK, and getting anything taken up is always a miracle to be celebrated. Except when it takes so long to come to market that they shouldnt have bothered. This is why open source is so much better as a way of doing tech transfer. If you have something good, a patch, a test and the ability to argue your case, it can be in the code tree in a week, and in people's hands the next day, in mainstream distros within a month or two.
Whereas MS Research? Vista took 5 years. Every new idea in the last three of those years will have been postponed to its successor. So the lag between an idea and product is 3-5 years, compared to 3-5 weeks.
-steve
I actually think for consumers, a media centre PC with Vista Ultimate is a very nice toy. good gaming, good media experience within the DRM sandbox they have constructed (one that seemingly excludes Zune, which can only be good). For that reason, its a real screwup to have missed christmas, which would have been the key time to launch it at the masses.
For enterprise, there are some bits that appeal. Bitlocker could be good for laptop security, though TPMs can be used to partially encrypt bits of an NTFS drive today. (they can keep the key, but all the file metadata is public, along with the swap and hibernate files). The new power management modes may bring the OS up to par with MacOS X, which again is good for laptops, as will be bluetooth support (finally!) and better WLAN. all of these are laptop features, which is important. In the F500 business world, the laptop is the primary machine. It costs the same, its support costs are higher, but the extra 2-5 hours of employee work you get per week soon pays for it.
All the corporates really need is a version of XP with disk security, decent bluetooth and power management, and good mobile networking, plus office apps that dont handle going off the LAN so badly, and a web browser that is not so prone to buffer overflows. In my F100 day job, firefox has a really bad experience, because so many must-use websites were written about 8 years ago and have hard coded assumptions about IE and its scripting language built in. Opera works much better. And you know what? IE7 has a hard time too. To the extend that not only is the automatic update disabled, if the PC audit tools find you have installed it voluntarily, you get email telling you off. They are going to rework and retest their sites for IE7, and only then can we upgrade to IE7 or vista, which wont be for seven months minumum.
Its that cost of upgrading every web site, as well as every corporate app, that is going to make adopting vista expensive, even if the system management costs are less.
And Me? I'm really looking forward to the upgraded web sites, as they should be firefox compatible too. Then I can stop running WinXP in a VMWare image under ubuntu.
-Steve
When I worked at a consumer PC vendor, along with the normal hard to fix incidents "sound goes off on mechwarrior on level 7", which take a while to replicate, those early adult DVDs caused a lot of problems on the first DVD-enabled PCS. These where the generation-1 DVD players, trying to render everything on a Pentium II/233 or maybe even a 266 MHz unit. Those adult DVD with their multi viewpoint features, and the need of the audience to freeze frames, caused a lot of problems. After about six months there was a shelf full of adult disks that were hard to play, alongside mainstream content that had high sustained bandwith, funny audio channels, complex menus etc. Funnily enough, the engineers were always willing to test the adult disks on new hardware and device drivers; it was the most thorough support team I ever knew. It was like they had a deep fear of regression failures creeping in. Word 97 not printing to postscript printers, AOL dialup failing, these were low priority compared to the different viewpoints in the orgy scenes working.
-steve
Wow, YaST worked for you? Since Suse 10.1 and move to using mono behind the scenes, it's been a heap of instability. I dream of it downloading WGA...
As someone who works on Apache ant, yes, we like bugreps that are replicable, and prefer patches with tests.
:(
At the same time, we try and test our stuff, internally and externally. But the moment an x.0 release ships, we still get lots of bugreps. And you know why that is? Because when the x.0 release ships, a lot more people grab the app and use it. And unlike beta testers, these are not experienced developers. They are people who (in the Java context) dont know that the CLASSPATH env variable is a recipie for disaster, that you shouldnt have trailing backslashes or inner quotes in it. We have people whose Windows PC is an inconsistent mess and things just dont work on them. We get people who are running jpackaged and self-installed ant distros side by side, and get surprised that ant.sh delegates to jpackage installations, so the upgrade doesnt appear to take.
The issue is not that we dont beta test our software, it is that the testers, having a certain level of competence/experience, don't set up the apps in a pathologically bad way. Its not that the code doesn't work, it is that we cannot test all configurations, and that is what burns us.
Operating systems have the same problem only multiplied.
One thing I dont agree with is closing bugs unless they are fixed, or unless the team has made a WONTFIX decision. The troublespot is WORKSFORME, because, yes, that is the problem: code that works on some configurations and not others. There is a great ongoing bugrep in Eclipse, that says "LATER bugs get ignored", which is how that team works. Marking something as later doesnt just postpone the fix, it hides it. In Ant, we leave all bugs open until closed propery. Which is why we have 500+ bugreps right now, I guess
-Steve
Apache Ant dev team; Author "Ant in Action"
That's a really good summary. RHEL are in the top 3 contributors to the linux codebase, next to linus and IBM. Nobody else understands it as well, or (outside Linus), has the ability to effect change. That's important, because it means redhat are one of the few groups who can set a direction for linux that suits them.
With the JBoss purchase, they even have influence in many java projects. Not all; I suspect google sun and IBM have more. but they own hibernate and jboss, and have a presence in tomcat.
I'm busy moving from Suse to ubuntu, then from jboss to Sun glassfish. not for idelogical reasons, but for technical ones, like "they work really well"
Windows pricing is hard to compare, because its so variable. A copy of XP pro over the three year life of a PC is less than RHWS; but add on the cost of CAL licenses to network infrastructure (exchange, file servers) and the price of a PC rises. Same for server installs -the cost of the database scales with the number of users you want to support.
On the other hand, if you are a developer with an MSDN subscription, you get all of windows, office, much of visual studio for a flat rate, less than an AS subscription. Microsoft treats developers well.
But Fedora? To increase the value of the paid products, RH had to reduce the value of the free ones. Fedora achieves that. I dont see why we 'community' developers, whose of us who write the stuff that shipps with RHAS, have to suffer on an unstable distro with no support for products like VMWare. Here I am, working on things for no money, and yet RHEL still treat me as a leech who deserves low stability junk. If I want that, I'd use MS Vista
In car players are still alive. the CD as a distribution format for playback in those systems still works. Only now you can burn a disk on demand, from your collection.
Somewhere up in my attic I have a pile of legacy analog audio technology. a record player, the LPs, a VCR and its tapes, a cassette recorder. The most valuable item is my grandfather's tube-based AM radio from 1938; I may bring that down and use it as a PC speaker. The rest is just lost space. Which is the problem I have with CDs. What do you do with them? Unless you want to show off your legacy use case, they just collect in a box?
yes. Every post-CD format (minidisk, DAT, SACD, DVD-Audio) has been crippled by DRM that reduced the value of the format. CD is the best format for content because they can sell it through the legacy retail channel (shops), and because it is mostly DRM free. Now, if a record company embraced DRM-free content, at a reasonable encoding and price, things would be different. After all, its not exactly hard to go from DRM protected content to unprotected, so why do they bother? A willingness to give iTunes total control of the business?
If every CD only has two good tracks on it, then the real value of those two tracks is $10 each. The other eight are giveaways. The record companies want variable pricing so they can charge more for those two tracks. The alternative strategy: try and sell more hasn't occurred to them.