Like most engineers, my writing skills were underdeveloped when I finished my master thesis. Then I became a Ph. D. student and learning to write acceptable texts in english became a necessity. I was lucky to have a supervisor who was a) a good writer b) a good peer reviewer. I was doing various things wrong systematically. Once that is pointed out to you and you understand why, you can then make it a point to do it right, always.
I'm sure there's some problems left, I'm a non native english speaker and will never become as fluent as native speakers. Since then I've supervised students myself and seen them make the same mistakes as I did. Some of these people have Ph. D.'s now. I know several people who attended writing courses as well (which is usually a more structured form of peer review). Also there is plenty of material out there for self study. And of course learning by example works well to: read!
As for things to do in courses is of course organizing proper peer review for written stuff. That is hard and takes up time. This why engineers write so poorly: their teachers prefer not to spend to much time on such technicalities as writing. Many master thesises for example are extremely poorly written but still good enough, apparently, for the student to get a title.
The big pitfall for engineers is to assume that writing is a creative process. It isn't, it is mostly about applying techniques and testing sentences and paragraphs against simple grammar and stylistic rules. It's the content that requires the creativity, not the form. Writer's block is about having nothing to write about rather than being unable to write. Once you figure that out, it is just a matter of picking up enough knowledge so that you can apply it.
Swapping is the bottleneck. So remove the bottleneck. No seriously. The harddisk activity you are most likely to notice is memory swapping. Swapping can be disabled. Of course you run out of memory if you do that, so add more memory. I find that with 2GB no application ever complains of having not enough memory despite there being exactly 0MB of swap space. I run some pretty memory intensive stuff too. It turns out most of this stuff is designed to run well on systems with only 512-1024 MB (particularly games rarely use more, even if it is avaialable). That extra GB is cheaper than a new harddrive and if 2 is not enough make it 3 or 4. It's not like win32 processes can address more than 2GB anyway!
At least under windows, memory swapping is implemented very stupidly. Basically the system will spend (your) time swapping even when there's plenty of memory available. I've observed it swapping applications to disk with over 75% memory available. This causes all sorts of noticable delays when you try to actually use your system (e.g. switching from application A to application B). With 2GB available, windows should run out of excuses to swap but it will still swap.
Disabling swap space effectively stops this behavior. Especially on slow harddisks this means a huge performance improvement. Depending on your software you can do with much less memory. I've disabled swap space on machines with only 512MB which you are unlikely to exceed running just office type applications. In all cases that I did this the result was an immediate, noticable performance increase.
In case you do run out of memory, you get an out of memory error. I find that closing applications usually is a good solution. Much better than windows continuously wasting my time with unnecessary UI blocking harddisk activity. Anyway, given the low cost of memory, I'm very intolerant towards having my time wasted due to the fact that there's not enough.
From a business perspective it makes perfect sense. The hardware is an expense to the value that the OS adds. Or rather, the margins on the software are comparatively much higher. Without the OS it is just about the same hardware Dell sells (with a prettier case around it). Additionally the apple hardware is currently constraining the market size to a fraction of its full potential. Even modest adoption of mac os X on average PC hardware is highly attractive growth scenario.
So, the question is not why Apple should do this but why not? But then this Cringely seems to be rarely do much more than blurt out the obvious wrapped up in some inflamatory way. I'm sure the guy knows nothing more than we do.
That's what happens when you give decision power to idiots. A most unfortunate side effect of democracy. A few idiots in charge of billions can do a lot of dammage (e.g. the Bush government).
On the research side, politics and funding always dominate the research agenda. Our modern scientific tradition started somewhere around the renaissance period. Funding in those days came from the catholic church which only recently and grudgingly accepted some of the scientific findings of that era as true.
Now even Darwin is finding acceptance in some religious corners. Intelligent Design people are firmly rooted in the conservative, anti change, anti progress corner. These people are the descendants of those that claimed the earth was flat, the sun is revolving around the earth (logically only after the previous was accepted as true), and that some guy with a beard created the earth in seven days. This strange collection of pseudo scientist, ignorant peasants and religious nuts can do a lot of dammage through their political power. Not only can they do this but they do so on a regular basis.
For research agenda's this usually means that they are driven primarily by short term political issues. Getting long term research funding is hard and it requires good marketing skills from the researcher. For evolutionary researchers in the US this means being friendly to those intelligent design nuts that otherwise vote your research projects away with their wallet.
It's costing them in the sense that they now give away software or charge very little for their entry level products. Of course the entire market is growing exponentially so they probably made healthy revenues despite all this. But arguably, before JBoss their entry level prices were quite a bit steeper.
Not in some magical way but by design and without scripting, I refer you the subversion book for further details. Really there's a lot in that book you clearly don't know. Basically subversion can do anything that cvs does (that was a design goal) and much more (because it couldn't be designed back into CVS, really it is quite bad). It turns out that the much more part is actually quite valuable if you care about version management. Many CVS users don't quite know what they are missing. You seem to be a good example.
anyway to counter your example how about svn commit -m "really stupid change on a really big directory affecting 10000 files" r100 svn delete svn+ssh://foo/trunk/dir -m "oops" r101 svn copy -r99 svn+ssh://foo/trunk/dir svn+ssh://foo/trunk -m "copy back last good version of dir" r102
r100 breaks the system so you delete the change (all the files under dir), don't worry no bits are actually lost because this is a proper versioning system. r100 is there to stay, you can always go back to it. So r101 does not have dir at all which is probably not something you want. So copy back the last good version from r99 which results in r102 (including version history and without messing with a stupid attic). which results in r102 which is identical to r99, no information lost.
Of course you'd want to maybe examine the changes between r100 and r101 in more detail before you embark on your litlle undo mission so you do a svn diff -r100:101 svn+ssh://foo/trunk which gives you a complete patch for that change (i.e. all diffs for all files that were changed in that commit). Just figuring out this information using cvs is a lot of work because cvs has no concept of all files in a particular commit because it only commits individual files. If you are used to svn log, cvs log really sucks I can assure you.
If you decide later on that revision 100 wasn't so bad after all but needs more work on a separate branch you do svn copy -r100 svn+ssh://foo/trunk svn+ssh://foo/branches/newbranch -m "interesting new stuff".
This little scenario is something that is quite impossible in cvs. You might be able to do it manually but good luck if the commit affected 10000 files under dir. BTW. cvs users would probably not dare to do such a commit in the first place (and rightly so, cvs is likely to not handle this gracefully).
Indeed you'd probably end up cvs diffing your way through the directory, scavenging what little information cvs did bother to put in the commit message which of course conveniently not includes what other files were committed at the same time. Have fun on a repository where 20 other people are committing while you are fixing things on the repository. The odds are that somebody manages to land a commit before you can fix your mess.
Subversion protects you in two ways: commits are atomic (so you can actually commit 10000 files at once if you want to) and the above scenario is time limited mainly by your typing skill so you could land your bad fix and have it removed in under 30 seconds. Should the worst happen (some one commits another change before you undo), you can actually apply an inverse diff to the head. That is you, do a svn diff -r100:99 > patch and then apply the patch to head and commit the change, barring conflicts this procedure is quite foolproof.
If you are going to use version management, do not standardize on obsolete toolsets. RCS is very limited. CVS works around some of its restrictions. Both tools are fairly limited. I'm sure it was pretty cool to put your three files in RCS in the eighties but modern times require more modern tools like that. There might actually be changes in two files that are related that neither cvs nor rcs is capable of keeping track off.
So use subversion. It's dead easy to set up, there's plenty of tool support. You can commit changesets (or rather, each commit is a changeset), which especially in the case of configuration files is nice since configuration changes are rarely limited to just one file. Alternatively you can tag/branch working configurations. You can refer to specific revisions of the entire system rather than specific revisions of specific files. In cvs tagging and branching are the only way to keep together the right versions of the right files, in subversion the revisionnumber tells you everything you need to know.
I would say subversion is adequate for version management (not perfect). It fullfills the minimum requirements of a versioning system. CVS on the other hand is unreliable, not really suitable for non ascii content, prone to losing version history on normal development activity such as e.g. refactoring, unable to keep track of changes that concern multiple files and unable to keep track of normal file operations such as directory creation, moving and deletion. Just a few of the reasons why you should no longer use CVS if you care about such things as repository integrity, keeping track of changes in the repository, storing binary content, refactoring your code, etc. If you do care about these things, avoid CVS like the plague. But then why use versioning at all if you don't care about these things? CVS is better than nothing and and svn is way better than that.
Tortoise CVS on windows does a good job of hiding most of the uglyness to end users, it does not actually solve them and users might be tempted to think that storing e.g. large word documents in cvs is good practice. TortoiseSVN, a frontend for subversion similar to TortoiseCVS provides a similar frontend for subversion. Having used both, I would say that the svn version is more feature rich, more actively developed and has the advantage of running on top of a version system that doesn't have any of the problems listed above.
To get started with subversion locally, all you need is tortoise svn. Simply right click on a folder and create a repository and then right click on another folder to add its contents to the newly created repository (using the file:// protocol). If at a later stage you decide you want to share the repository over a network, you can do so with e.g. cygwin ssh + cygwin subversion using the svn+ssh:// protocol. The tortoise svn repository will actually be compatible with that. Alternatively you can set up an apache server (http:/// or svnserve (svn://). For small scale sharing, svn+ssh is easier since it requires nothing more than installing subversion, getting the ssh server going and managing ssh users (which in cygwin are normal windows users).
Due to subversions excellent support for normal file operations, you can actually have only one repository for multiple projects. For example apache projects share the same subversion repository.
Because it is so easy to set up, subversion is really suitable for working on a bunch of files together with others. I use a local repository for working on my work documents and small projects. I even put photo's in subversion. Having a central repository makes it easy to backup as well. This style of working with subversion is reasonably efficient for binary files where the binary diff from version to version is relatively small too. For example, I keep my word documents directory in svn. The total size of an export of this directory is 10 MB. The subversion repository on disk containing the full version history of these files is 5 MB. Compression + storing only binary diffs are the reason that the repository size on disk is actually less than an export of the head of the repository!
System administrators tend to want to restrict users. For example, I'm behind a proxy at work. This happens to interfere with my job sometimes so I work around it in ways that are worse then just opening up the proxy (like doing some of my work at home where I am in control over the network).
Users tend to walk around obstacles in their path. A usb stick is only one of the many ways users can work around restrictions. The solution is to enable them to do their thing. If allowing them to do their work conflicts with security the stupid solution is to prevent them from doing so. Instant messaging clients are a good example. Many corporations block this kind of software. Many employees work around this. For example by doing it from their home pc. Before you know it they're sending work related files to colleagues over msn from home over an unsecured dialup connection on a home pc loaded with spyware. That's actually much worse then selectively allowing IM type communication over the corporate network. Corporate networks are full of these social hack enabling type holes.
"With rare exceptions, porting hassles between UNIX systems are long forgotten."
Yeah right. We're down to complaining about porting apps between versions of the same distribution of linux and here's a guy claiming that porting hassles between UNIX systems ar long forgotten. Come on, you can't claim that with a straight face even if you are working for Microsoft.
It's not like your 1993 binary of wolfenstein will work out of the box on win XP but the chances of binaries from that era doing something are a lot higher than say running any unix binary from that era on the latest Red Hat/Debian/Whatever. Compiling with the latest GCC almost guarantees running into porting issues, never mind the particular OS you are doing that on. Probably hello world works fine but anything doing something less trivial is likely to not work at all. Generally problems increase as you (necessarily for non trivial stuff) depend on libraries not part of any of the Unix standards. For example, these standards do not cover anything related to graphics or user interfaces so you're fucked if your app is non trivial enough to include a (G)UI.
Good example of lack of progress. Just look at the list of who's not participating and the list of stuff not covered by this. Just the zillionth attempt to fix the wrong issues.
We don't need some independent external group doing this kind of thing but representatives of all main distributions getting together to specify once and for all what is and is not desktop linux. None of them is even interested in doing that. As long as they don't do that there will be no desktop linux.
He certainly is right. He just highlighted the #1 problem (by far) that has been keeping desktop linux back. It's been obvious to me for years that this is a big problem. I've argued so on various occasions. Not only is it difficult for hardware vendors to offer linux support, it is equally hard to offer meaningful software support. Linux distributions are very different from each other and on top of that have poor compatibility with earlier versions of the same distribution (try installing some potato debs on the latest debian testing). Effectively, testing for two different distributions is like testing for an entirely different OS. The most succesful strategy for getting software to work across distributions is to minimize dependencies on the OS. For some popular linux applications that includes bundling application specific libraries for such things as String manipulation, displaying GUIs. I'm of course talking about the two linux desktop successtories firefox and openoffice (both of which have more windows than linux users btw.).
When people are talking about 'desktop linux' it is totally unclear what subset of versions of specific packages they are referring to. Effectively, there's no such thing as desktop linux. Desktop linux is not something currently available. There's lots of distributions claiming to offer it, only they're all offering a different implementation of that concept. The whole distribution based thinking ensures that the problem will continue to exist. As long gentoo, red-hat, suse, debian and ubuntu (and the many niche distributions organized around these) continue to not work together on simple cross distribution standard behavior the situation will not improve.
There's of course some limited effort progressing at a snail pace (has been going on for more than half a decade) on standardizing some things like embedable objects, printing, widget libraries, etc. You know the stuff that worked pretty ok in windows 3.0. The standardization efforts do not extend to such important things as infrastructure for installing distribution independent binary software packages (such as for example drivers, games). Consequently, such packages are rarely available from other channels than the distribution provider. If it's not in the distribution repository, good luck getting it to work (mileage varies wildly from 'might work' to 'will break various other software you already have').
Essentially your company is asking you to make a lot of commitment and invest a lot of effort. The least they could do is finance the whole thing.
I recently moved to Finland (from the Netherlands). This was my third international move and I have some experience in this kind of thing by now. Basicallly you are going to run into a number of problems:
- finding an appartment in a place you don't know to which you have limited access until you move there. Sort of a chicken egg problem. You'll need all the help you can get. - You have a busy job. That means little time to do stuff like organizing a move; finding an appartment; etc. - bureaucracy & culture. I had to deal with the additional burden of strange new culture, lots of bureaucratic stuff I knew nothing about, etc. I imagine things like taxes, insurance, etc. work differently on the west coast than on the east coast. Having someone to advice you on that is quite handy. - You are going to waste a lot of energy on tedious things like making sure the phone is disconnected at the right time and getting a new one in your new place. - there's going to be either a period where you either have two houses or a period where you have no house at all and are effectively homeless. The latter case can be quite expensive (hotel) and the first is as well. - You are going to be offline for a few weeks (you read/. so you must care about that).
Things your employer can help you with: - getting you settled in. Depending on your status in this company this can actually range from doing just about anything for you (really, there are companies specialized in this kind of thing, think door to door service) to paying for a moving company (that would be the bare minimum). In my case, I got a pretty good deal which ensured minimal stress during the whole procedure. I had help finding an apartment, i was escorted to the various government offices, I spent two weeks in a very good hotel, my flight was paid for, etc. Also I was compensated financially for the moving cost. - Giving you some time off around the time of moving. Not having to dive into work related stuff for a few weeks gives you some opportunity to settle in properly. Just getting your stuff packed - Financial security. You are making a commitment that goes beyond paying for the moving company. It is not strange to ask for compensation for that. You are giving up friends, family. You'll be making lots of trips back and forth visiting them. You are maybe selling your house, car etc. You'll be buying new stuff for your new place, etc.
Of course it all depends on how badly they want you. It could be like move here now or you're fired or it could be like we would really like you to move here and continue working for us and hey here's a nice raise in salary. Bottom line is that full compensation is going to cost the company quite a bit. Are they willing to do that? Rule of thumb is that you have to ask for it before you agree, not after. It's part of the whole deal.
It's not like 300GB disks are expensive. Buy a few of them, put them in a raid. Don't back up to tape but make incremental backups to a separate disk array. Problem solved, hw cost a few thousand $ at most.
And remember, you're there to support your users, not the other way around. It's not up to you to determine what's important or not. If the guy says his four year mail is important, it is and it's your job to ensure it is accessible and securely backed up every night.
Everybody here seems to be seeing conspiracies here. Database licenses are increasingly less important to Oracle as their revenues from application servers and services are growing (and already larger than revenue from database licenses). By owning relevant products that cover the full spectrum from embedded databases to full blown enterprise databases they ensure that they are the guys to talk to when considering doing anything involving database storage.
Oracle now has in its productline: - several databases - application servers - IDEs - ERP software
All of which serve a single purpose: selling services to enterprises. License revenue is nice but peanuts compared to the services component of the bill. The low end of the market is where the growth is and mysql dominates that market. Oracle's latest aqcuisitions are merely their entry ticket to this market. They're not there to destroy the market but to take it over and make sure that they can sell their services in that market as well.
That's just a difficult way of reasoning away the need for a god (a vision I agree with).
Historically, god (or gods) have been used to explain stuff that was not understood. In some levels of the world with very poor understanding of just about anything this includes almost every aspect of life. In the parts of the world where people have accomplished such things as e.g. figuring out that the earth is in fact not flat or putting people on that yellow disc in the sky, people tend to take a more sophisticated view. But even there most people tend to have a very blurred understanding of the things around them relying on a scientific and religious elite (or sometimes both!) to help them explain everyday life.
Darwin did an interesting thing. He developed a theory that helped explain a lot of biological phenomena that up until then had been the domain of wildly speculative stories involving apples, clay etc. This article is about some of the supporters of these silly fairy tales grudgingly but not whole hartedly admitting that Darwin might have been right about a thing or two (which given all accomplishments dependent on Darwin's theory is getting kind of hard) and trying to argue the compatibility of that particular theory with their own theory involving apples and clay. Silly of course but the confused masses are easy to convince.
Sadly intelligence is no longer driving our own evolution. The stupid are reproducing like rabbits.
Vista won't be out until the end of the year. So I don't see the problem. This new standard is not going to be supported by anything else than vista. Aside from a bunch of tweakers, the only way people will get Vista is by buying a new PC. That's why nvidia and ati are not bothering to put useless hardware on their current boards. I'm sure that if there is any market demand for this standard, there will be some compatible hardware by the time Vista launches.
Of course the big question is weather this standard will work at all. If you take a step back and look at what the industry is doing, you see a lot of vertical stacks of technology with none of them well positioned for long term success. IMHO neither blue ray or hddvd is going to have any long term relevance. The HDCP standard will add to this problem since it will complicate and slow adoption of the new technology. That in turn means lower demand for HD content.
If you look at the long term, the only relevant distribution channel for any digital content is online distribution. Once the industry decides that online distribution is the way forward, the whole mess of vertical technology will more or less automatically ensure that any technology which restricts market share will be extemely unpopular with consumers and, ironically, content distributers. Why sell onlince content to only 1% of the market with compliant hardware when you can sell to 100% of the market with good enough hardware?
The first company who gets this right will make lots of money real fast.
Agreed. Except the changes that need to take place are probably much more like netscape 4 vs mozilla 1.0 than mozilla 1.x vs Firefox 1.0. Essentially firefox was only about the UI whereas mozilla was a complete rewrite of netscape 4.x.
Open office consists of close to 10 million lines of code. Much of it is star office legacy code that very few people understand. Because of this legacy, feature development happens at a snailpace and the UI looks like shit.
A complete rewrite is not a realistic option for Sun. Doing so certainly killed Netscape and their product was a much smaller software product. A full rewrite would take the best part of this decade. A more realistic strategy would be to gradually replace the worst parts of the system. Identify the problematic components and fix or redesign them. Sure the UI sucks but it sort of works. The 2.0 development even made it look semi native on some platforms. Deoupling the components of the suit means that you can evolve them (or kill them off) seperately.
Actually you are quite right. The big problem in the linux community is that they are not really open to this kind of criticism. People have been saying for years that UI sucks and for years it has continued to suck.
Lets face it, the GIMP UI is pretty bad. It is going to take some major rearchitecting to fix that. One reason why that won't happen is that the people who are supposed to do that obviously don't get it (just look at the current UI...).
I have a canon powershot A40, which is now five years old. It's a 2 megapixel camera and was a pretty good thing for the time. Over the years I've learned to make pictures with it that I can fix in photoshop. The ccd sensor provides plenty of detail but the camera software will seriously affect image quality unless you intervene. Images tend to be overexposed, slightly blueish with too much contrast. Now underexposure is really easy to fix. Overexposure on the other hand means white areas in the picture with no detail whatsoever. Similarly, contrast improving algorithms are lossy: they actually throw away a lot of detail. So what I do is shoot pictures with the manual mode that are a bit underexposed. Then I load them up in photoshop and fix them. First I fix the levels. Then, if necessary, I add a curves layer to boost shadow detail a bit. Finally I adjust the color balance to compensate for the blue (especially shots with snow in them). I try not to use the brightness and contrast controls as both are lossy (just look at what happens to the histogram if you use them).
Good camera software should not get in the way of this process. Unfortunately, many cameras do get in the way deliberately. Basically all the low end cameras try to 'fix' the photo in the camera so the users don't have to. The end result with the better ones is not that bad but you'd be able to get much more out of the photo if you'd have the raw material in photoshop. That's why all the highend cameras support the raw format.
What I would like is a compact high end camera. Even the pros adit it: they shoot most shots with some cheap compact camera instead of the bulky SLR they leave home most of the time. I don't have an SLR and I don't think I want one. I do want some of the features though. My compact high end camera of choice would sport great image quality, raw format support, full control over everything (and some convenient defaults for most) and good camera software that does not throw away detail on my behalf.
I'm seriously considering the canon S80. If it had raw support I would have bought it already.
Particularly because running the two on the same platform is a bit hard. Mono runs on windows, sort of, if you want to. But forget about getting any of the mono tools working on windows. And lets be fair, most people currently developing on visual studio are unlikely to run anything else than windows. So even if they wanted to, there's no way for them to even try out monodevelop short of installing linux.
I have to admit Sun seems to have some positive swing lately. They're selling some hardware again. Their open source policy finally seems to be bringing in new customers, after years of shooting themselves in the foot with weird licenses, inconsistent marketing and plain corporate stupidity affecting all their business units.
Their strategy of the past years has been very ineffective. Java has become a multi billion industry, Sun has invested hugely in it but they have failed to cash in on it directly (I suspect they barely break even). Also it hasn't driven hardware sales that much. Their main competitor on the other hand seems to have a very succesful Java strategy. IBM is leading the way in application servers, IDEs (eclipse, rational rose) and middleware with basically the whole industry eating out of their hand, including most JCP specifications committees. On top of that they also sell the hardware & support to go with the software. IBM loves Java!
Then the whole x86 solaris thing has come a long way too (from 'hey it's free now but we support it for a fee', 'oh wait we don't do that anymore' to 'oh well lets open source the whole thing and forget about it' to finally 'hey x86 solaris is really important to us'). I mean, what do they want?
Sparc sales have been a disaster for the past years with people basically favoring IBM power and x86 with linux. They may be laughing at intel for itanium but they have one thing in common with intel: x86 is driving sales for both of them. Sun has a few next generation architectures on the shelves which no doubt they are going to try to sell. These chips had better be way better than the competition (and their mediocre current offerings) or otherwise whatever spin they put on it won't work. Personally I'm not convinced yet. Their two operating systems (solaris/linux) & two architectures (x86/sparc4 aka niagra) policy is going to continue to confuse people. Sun seems to think the combinations can coexist without affecting each others marketshare. I don't. Confused customers will look to IBM and others.
To me Sun still is a company in trouble. Maybe a few of their business units are recovering (finally) but that still leaves large parts of the company not performing very well.
Gates is a bit of an oddball. I can't think of anything particularly brilliant he's ever done and yet he's the richest guy on the planet so he must be doing something right. Arguably making a deal was with IBM was the smartest thing he's ever done (and the stupidest thing IBM has ever done). But from a technology perspective, nothing exciting happened at the time. I mean DOS was a shitty system, even at the time it was introduced.
Jobs on the other hand has always associated himself with cool stuff though none of it can really be attributed to him. He was just sort of there at the right moment, surrounded by brilliant people doing really great stuff.
If I'd have to pick one it would be Jobs. Mainly because I like people capable of thinking out of the box. There's too few of those in this world. And Jobs has certainly proved that he's capable of that. Gates on the other hand... well need I remind anyone of his 1995 visionairy masterpiece in which he managed to almost totally ignore the internet? To this very day the guy rarely says or does anything interesting. He's sitting on a huge pile of cash but other than giving it away to charity (which is good) he's not doing anything interesting with it. Somehow, I think Steve Jobs would never be able to just sit on a pile of cash like that. He'd be itching to spend it on something, anything.
XHTML is a good illustration why it is a good rule to never finalize a standard without a usable reference implementation. Right now it's just a boring document. I suspect the people who are supposed to implement it have barely read it. And frankly, I don't think they will.
XHTML 2.0 is part of a fictional roadmap that was tossed around late last century and then was discarded. Microsoft decided to sort of stop developing their rendering engine and Netscape was sold to AOL and eventually was liquidated. Microsoft is still pretty happy to not do any significant new standards support in their upcoming browser. Meanwhile the mozilla people, opera people and konqueror/safari people are working outside the w3c to get some real standards work done: like this decade! Html 5.0 is something I can see happen, xhtml 2.0 was dead. on arrival (did they ever bother to make it a recommended spec?). As a matter of fact, XHTML 1.1 is as unlikely to ever be a relevant standard as well. And that one would be relatively easy to implement for browser developers.
Slashdot Mirror
Like most engineers, my writing skills were underdeveloped when I finished my master thesis. Then I became a Ph. D. student and learning to write acceptable texts in english became a necessity. I was lucky to have a supervisor who was a) a good writer b) a good peer reviewer. I was doing various things wrong systematically. Once that is pointed out to you and you understand why, you can then make it a point to do it right, always.
I'm sure there's some problems left, I'm a non native english speaker and will never become as fluent as native speakers. Since then I've supervised students myself and seen them make the same mistakes as I did. Some of these people have Ph. D.'s now. I know several people who attended writing courses as well (which is usually a more structured form of peer review). Also there is plenty of material out there for self study. And of course learning by example works well to: read!
As for things to do in courses is of course organizing proper peer review for written stuff. That is hard and takes up time. This why engineers write so poorly: their teachers prefer not to spend to much time on such technicalities as writing. Many master thesises for example are extremely poorly written but still good enough, apparently, for the student to get a title.
The big pitfall for engineers is to assume that writing is a creative process. It isn't, it is mostly about applying techniques and testing sentences and paragraphs against simple grammar and stylistic rules. It's the content that requires the creativity, not the form. Writer's block is about having nothing to write about rather than being unable to write. Once you figure that out, it is just a matter of picking up enough knowledge so that you can apply it.
Swapping is the bottleneck. So remove the bottleneck. No seriously. The harddisk activity you are most likely to notice is memory swapping. Swapping can be disabled. Of course you run out of memory if you do that, so add more memory. I find that with 2GB no application ever complains of having not enough memory despite there being exactly 0MB of swap space. I run some pretty memory intensive stuff too. It turns out most of this stuff is designed to run well on systems with only 512-1024 MB (particularly games rarely use more, even if it is avaialable). That extra GB is cheaper than a new harddrive and if 2 is not enough make it 3 or 4. It's not like win32 processes can address more than 2GB anyway!
At least under windows, memory swapping is implemented very stupidly. Basically the system will spend (your) time swapping even when there's plenty of memory available. I've observed it swapping applications to disk with over 75% memory available. This causes all sorts of noticable delays when you try to actually use your system (e.g. switching from application A to application B). With 2GB available, windows should run out of excuses to swap but it will still swap.
Disabling swap space effectively stops this behavior. Especially on slow harddisks this means a huge performance improvement. Depending on your software you can do with much less memory. I've disabled swap space on machines with only 512MB which you are unlikely to exceed running just office type applications. In all cases that I did this the result was an immediate, noticable performance increase.
In case you do run out of memory, you get an out of memory error. I find that closing applications usually is a good solution. Much better than windows continuously wasting my time with unnecessary UI blocking harddisk activity. Anyway, given the low cost of memory, I'm very intolerant towards having my time wasted due to the fact that there's not enough.
From a business perspective it makes perfect sense. The hardware is an expense to the value that the OS adds. Or rather, the margins on the software are comparatively much higher. Without the OS it is just about the same hardware Dell sells (with a prettier case around it). Additionally the apple hardware is currently constraining the market size to a fraction of its full potential. Even modest adoption of mac os X on average PC hardware is highly attractive growth scenario.
So, the question is not why Apple should do this but why not? But then this Cringely seems to be rarely do much more than blurt out the obvious wrapped up in some inflamatory way. I'm sure the guy knows nothing more than we do.
That's what happens when you give decision power to idiots. A most unfortunate side effect of democracy. A few idiots in charge of billions can do a lot of dammage (e.g. the Bush government).
On the research side, politics and funding always dominate the research agenda. Our modern scientific tradition started somewhere around the renaissance period. Funding in those days came from the catholic church which only recently and grudgingly accepted some of the scientific findings of that era as true.
Now even Darwin is finding acceptance in some religious corners. Intelligent Design people are firmly rooted in the conservative, anti change, anti progress corner. These people are the descendants of those that claimed the earth was flat, the sun is revolving around the earth (logically only after the previous was accepted as true), and that some guy with a beard created the earth in seven days. This strange collection of pseudo scientist, ignorant peasants and religious nuts can do a lot of dammage through their political power. Not only can they do this but they do so on a regular basis.
For research agenda's this usually means that they are driven primarily by short term political issues. Getting long term research funding is hard and it requires good marketing skills from the researcher. For evolutionary researchers in the US this means being friendly to those intelligent design nuts that otherwise vote your research projects away with their wallet.
It's costing them in the sense that they now give away software or charge very little for their entry level products. Of course the entire market is growing exponentially so they probably made healthy revenues despite all this. But arguably, before JBoss their entry level prices were quite a bit steeper.
Not in some magical way but by design and without scripting, I refer you the subversion book for further details. Really there's a lot in that book you clearly don't know. Basically subversion can do anything that cvs does (that was a design goal) and much more (because it couldn't be designed back into CVS, really it is quite bad). It turns out that the much more part is actually quite valuable if you care about version management. Many CVS users don't quite know what they are missing. You seem to be a good example.
anyway to counter your example how about
svn commit -m "really stupid change on a really big directory affecting 10000 files"
r100
svn delete svn+ssh://foo/trunk/dir -m "oops"
r101
svn copy -r99 svn+ssh://foo/trunk/dir svn+ssh://foo/trunk -m "copy back last good version of dir"
r102
r100 breaks the system
so you delete the change (all the files under dir), don't worry no bits are actually lost because this is a proper versioning system. r100 is there to stay, you can always go back to it.
So r101 does not have dir at all which is probably not something you want.
So copy back the last good version from r99 which results in r102 (including version history and without messing with a stupid attic).
which results in r102 which is identical to r99, no information lost.
Of course you'd want to maybe examine the changes between r100 and r101 in more detail before you embark on your litlle undo mission so you do a svn diff -r100:101 svn+ssh://foo/trunk which gives you a complete patch for that change (i.e. all diffs for all files that were changed in that commit). Just figuring out this information using cvs is a lot of work because cvs has no concept of all files in a particular commit because it only commits individual files. If you are used to svn log, cvs log really sucks I can assure you.
If you decide later on that revision 100 wasn't so bad after all but needs more work on a separate branch you do svn copy -r100 svn+ssh://foo/trunk svn+ssh://foo/branches/newbranch -m "interesting new stuff".
This little scenario is something that is quite impossible in cvs. You might be able to do it manually but good luck if the commit affected 10000 files under dir. BTW. cvs users would probably not dare to do such a commit in the first place (and rightly so, cvs is likely to not handle this gracefully).
Indeed you'd probably end up cvs diffing your way through the directory, scavenging what little information cvs did bother to put in the commit message which of course conveniently not includes what other files were committed at the same time. Have fun on a repository where 20 other people are committing while you are fixing things on the repository. The odds are that somebody manages to land a commit before you can fix your mess.
Subversion protects you in two ways: commits are atomic (so you can actually commit 10000 files at once if you want to) and the above scenario is time limited mainly by your typing skill so you could land your bad fix and have it removed in under 30 seconds. Should the worst happen (some one commits another change before you undo), you can actually apply an inverse diff to the head. That is you, do a svn diff -r100:99 > patch and then apply the patch to head and commit the change, barring conflicts this procedure is quite foolproof.
If you are going to use version management, do not standardize on obsolete toolsets. RCS is very limited. CVS works around some of its restrictions. Both tools are fairly limited. I'm sure it was pretty cool to put your three files in RCS in the eighties but modern times require more modern tools like that. There might actually be changes in two files that are related that neither cvs nor rcs is capable of keeping track off.
So use subversion. It's dead easy to set up, there's plenty of tool support. You can commit changesets (or rather, each commit is a changeset), which especially in the case of configuration files is nice since configuration changes are rarely limited to just one file. Alternatively you can tag/branch working configurations. You can refer to specific revisions of the entire system rather than specific revisions of specific files. In cvs tagging and branching are the only way to keep together the right versions of the right files, in subversion the revisionnumber tells you everything you need to know.
I would say subversion is adequate for version management (not perfect). It fullfills the minimum requirements of a versioning system. CVS on the other hand is unreliable, not really suitable for non ascii content, prone to losing version history on normal development activity such as e.g. refactoring, unable to keep track of changes that concern multiple files and unable to keep track of normal file operations such as directory creation, moving and deletion. Just a few of the reasons why you should no longer use CVS if you care about such things as repository integrity, keeping track of changes in the repository, storing binary content, refactoring your code, etc. If you do care about these things, avoid CVS like the plague. But then why use versioning at all if you don't care about these things? CVS is better than nothing and and svn is way better than that.
Tortoise CVS on windows does a good job of hiding most of the uglyness to end users, it does not actually solve them and users might be tempted to think that storing e.g. large word documents in cvs is good practice. TortoiseSVN, a frontend for subversion similar to TortoiseCVS provides a similar frontend for subversion. Having used both, I would say that the svn version is more feature rich, more actively developed and has the advantage of running on top of a version system that doesn't have any of the problems listed above.
To get started with subversion locally, all you need is tortoise svn. Simply right click on a folder and create a repository and then right click on another folder to add its contents to the newly created repository (using the file:// protocol). If at a later stage you decide you want to share the repository over a network, you can do so with e.g. cygwin ssh + cygwin subversion using the svn+ssh:// protocol. The tortoise svn repository will actually be compatible with that. Alternatively you can set up an apache server (http:/// or svnserve (svn://). For small scale sharing, svn+ssh is easier since it requires nothing more than installing subversion, getting the ssh server going and managing ssh users (which in cygwin are normal windows users).
Due to subversions excellent support for normal file operations, you can actually have only one repository for multiple projects. For example apache projects share the same subversion repository.
Because it is so easy to set up, subversion is really suitable for working on a bunch of files together with others. I use a local repository for working on my work documents and small projects. I even put photo's in subversion. Having a central repository makes it easy to backup as well. This style of working with subversion is reasonably efficient for binary files where the binary diff from version to version is relatively small too. For example, I keep my word documents directory in svn. The total size of an export of this directory is 10 MB. The subversion repository on disk containing the full version history of these files is 5 MB. Compression + storing only binary diffs are the reason that the repository size on disk is actually less than an export of the head of the repository!
System administrators tend to want to restrict users. For example, I'm behind a proxy at work. This happens to interfere with my job sometimes so I work around it in ways that are worse then just opening up the proxy (like doing some of my work at home where I am in control over the network).
Users tend to walk around obstacles in their path. A usb stick is only one of the many ways users can work around restrictions. The solution is to enable them to do their thing. If allowing them to do their work conflicts with security the stupid solution is to prevent them from doing so. Instant messaging clients are a good example. Many corporations block this kind of software. Many employees work around this. For example by doing it from their home pc. Before you know it they're sending work related files to colleagues over msn from home over an unsecured dialup connection on a home pc loaded with spyware. That's actually much worse then selectively allowing IM type communication over the corporate network. Corporate networks are full of these social hack enabling type holes.
"With rare exceptions, porting hassles between UNIX systems are long forgotten."
Yeah right. We're down to complaining about porting apps between versions of the same distribution of linux and here's a guy claiming that porting hassles between UNIX systems ar long forgotten. Come on, you can't claim that with a straight face even if you are working for Microsoft.
It's not like your 1993 binary of wolfenstein will work out of the box on win XP but the chances of binaries from that era doing something are a lot higher than say running any unix binary from that era on the latest Red Hat/Debian/Whatever. Compiling with the latest GCC almost guarantees running into porting issues, never mind the particular OS you are doing that on. Probably hello world works fine but anything doing something less trivial is likely to not work at all. Generally problems increase as you (necessarily for non trivial stuff) depend on libraries not part of any of the Unix standards. For example, these standards do not cover anything related to graphics or user interfaces so you're fucked if your app is non trivial enough to include a (G)UI.
Good example of lack of progress. Just look at the list of who's not participating and the list of stuff not covered by this. Just the zillionth attempt to fix the wrong issues.
We don't need some independent external group doing this kind of thing but representatives of all main distributions getting together to specify once and for all what is and is not desktop linux. None of them is even interested in doing that. As long as they don't do that there will be no desktop linux.
He certainly is right. He just highlighted the #1 problem (by far) that has been keeping desktop linux back. It's been obvious to me for years that this is a big problem. I've argued so on various occasions. Not only is it difficult for hardware vendors to offer linux support, it is equally hard to offer meaningful software support. Linux distributions are very different from each other and on top of that have poor compatibility with earlier versions of the same distribution (try installing some potato debs on the latest debian testing). Effectively, testing for two different distributions is like testing for an entirely different OS. The most succesful strategy for getting software to work across distributions is to minimize dependencies on the OS. For some popular linux applications that includes bundling application specific libraries for such things as String manipulation, displaying GUIs. I'm of course talking about the two linux desktop successtories firefox and openoffice (both of which have more windows than linux users btw.).
When people are talking about 'desktop linux' it is totally unclear what subset of versions of specific packages they are referring to. Effectively, there's no such thing as desktop linux. Desktop linux is not something currently available. There's lots of distributions claiming to offer it, only they're all offering a different implementation of that concept. The whole distribution based thinking ensures that the problem will continue to exist. As long gentoo, red-hat, suse, debian and ubuntu (and the many niche distributions organized around these) continue to not work together on simple cross distribution standard behavior the situation will not improve.
There's of course some limited effort progressing at a snail pace (has been going on for more than half a decade) on standardizing some things like embedable objects, printing, widget libraries, etc. You know the stuff that worked pretty ok in windows 3.0. The standardization efforts do not extend to such important things as infrastructure for installing distribution independent binary software packages (such as for example drivers, games). Consequently, such packages are rarely available from other channels than the distribution provider. If it's not in the distribution repository, good luck getting it to work (mileage varies wildly from 'might work' to 'will break various other software you already have').
Essentially your company is asking you to make a lot of commitment and invest a lot of effort. The least they could do is finance the whole thing.
/. so you must care about that).
I recently moved to Finland (from the Netherlands). This was my third international move and I have some experience in this kind of thing by now. Basicallly you are going to run into a number of problems:
- finding an appartment in a place you don't know to which you have limited access until you move there. Sort of a chicken egg problem. You'll need all the help you can get.
- You have a busy job. That means little time to do stuff like organizing a move; finding an appartment; etc.
- bureaucracy & culture. I had to deal with the additional burden of strange new culture, lots of bureaucratic stuff I knew nothing about, etc. I imagine things like taxes, insurance, etc. work differently on the west coast than on the east coast. Having someone to advice you on that is quite handy.
- You are going to waste a lot of energy on tedious things like making sure the phone is disconnected at the right time and getting a new one in your new place.
- there's going to be either a period where you either have two houses or a period where you have no house at all and are effectively homeless. The latter case can be quite expensive (hotel) and the first is as well.
- You are going to be offline for a few weeks (you read
Things your employer can help you with:
- getting you settled in. Depending on your status in this company this can actually range from doing just about anything for you (really, there are companies specialized in this kind of thing, think door to door service) to paying for a moving company (that would be the bare minimum). In my case, I got a pretty good deal which ensured minimal stress during the whole procedure. I had help finding an apartment, i was escorted to the various government offices, I spent two weeks in a very good hotel, my flight was paid for, etc. Also I was compensated financially for the moving cost.
- Giving you some time off around the time of moving. Not having to dive into work related stuff for a few weeks gives you some opportunity to settle in properly. Just getting your stuff packed
- Financial security. You are making a commitment that goes beyond paying for the moving company. It is not strange to ask for compensation for that. You are giving up friends, family. You'll be making lots of trips back and forth visiting them. You are maybe selling your house, car etc. You'll be buying new stuff for your new place, etc.
Of course it all depends on how badly they want you. It could be like move here now or you're fired or it could be like we would really like you to move here and continue working for us and hey here's a nice raise in salary. Bottom line is that full compensation is going to cost the company quite a bit. Are they willing to do that? Rule of thumb is that you have to ask for it before you agree, not after. It's part of the whole deal.
It's not like 300GB disks are expensive. Buy a few of them, put them in a raid. Don't back up to tape but make incremental backups to a separate disk array. Problem solved, hw cost a few thousand $ at most.
And remember, you're there to support your users, not the other way around. It's not up to you to determine what's important or not. If the guy says his four year mail is important, it is and it's your job to ensure it is accessible and securely backed up every night.
Everybody here seems to be seeing conspiracies here. Database licenses are increasingly less important to Oracle as their revenues from application servers and services are growing (and already larger than revenue from database licenses). By owning relevant products that cover the full spectrum from embedded databases to full blown enterprise databases they ensure that they are the guys to talk to when considering doing anything involving database storage.
Oracle now has in its productline:
- several databases
- application servers
- IDEs
- ERP software
All of which serve a single purpose: selling services to enterprises. License revenue is nice but peanuts compared to the services component of the bill. The low end of the market is where the growth is and mysql dominates that market. Oracle's latest aqcuisitions are merely their entry ticket to this market. They're not there to destroy the market but to take it over and make sure that they can sell their services in that market as well.
That's just a difficult way of reasoning away the need for a god (a vision I agree with).
Historically, god (or gods) have been used to explain stuff that was not understood. In some levels of the world with very poor understanding of just about anything this includes almost every aspect of life. In the parts of the world where people have accomplished such things as e.g. figuring out that the earth is in fact not flat or putting people on that yellow disc in the sky, people tend to take a more sophisticated view. But even there most people tend to have a very blurred understanding of the things around them relying on a scientific and religious elite (or sometimes both!) to help them explain everyday life.
Darwin did an interesting thing. He developed a theory that helped explain a lot of biological phenomena that up until then had been the domain of wildly speculative stories involving apples, clay etc. This article is about some of the supporters of these silly fairy tales grudgingly but not whole hartedly admitting that Darwin might have been right about a thing or two (which given all accomplishments dependent on Darwin's theory is getting kind of hard) and trying to argue the compatibility of that particular theory with their own theory involving apples and clay. Silly of course but the confused masses are easy to convince.
Sadly intelligence is no longer driving our own evolution. The stupid are reproducing like rabbits.
Vista won't be out until the end of the year. So I don't see the problem. This new standard is not going to be supported by anything else than vista. Aside from a bunch of tweakers, the only way people will get Vista is by buying a new PC. That's why nvidia and ati are not bothering to put useless hardware on their current boards. I'm sure that if there is any market demand for this standard, there will be some compatible hardware by the time Vista launches.
Of course the big question is weather this standard will work at all. If you take a step back and look at what the industry is doing, you see a lot of vertical stacks of technology with none of them well positioned for long term success. IMHO neither blue ray or hddvd is going to have any long term relevance. The HDCP standard will add to this problem since it will complicate and slow adoption of the new technology. That in turn means lower demand for HD content.
If you look at the long term, the only relevant distribution channel for any digital content is online distribution. Once the industry decides that online distribution is the way forward, the whole mess of vertical technology will more or less automatically ensure that any technology which restricts market share will be extemely unpopular with consumers and, ironically, content distributers. Why sell onlince content to only 1% of the market with compliant hardware when you can sell to 100% of the market with good enough hardware?
The first company who gets this right will make lots of money real fast.
Agreed. Except the changes that need to take place are probably much more like netscape 4 vs mozilla 1.0 than mozilla 1.x vs Firefox 1.0. Essentially firefox was only about the UI whereas mozilla was a complete rewrite of netscape 4.x.
Open office consists of close to 10 million lines of code. Much of it is star office legacy code that very few people understand. Because of this legacy, feature development happens at a snailpace and the UI looks like shit.
A complete rewrite is not a realistic option for Sun. Doing so certainly killed Netscape and their product was a much smaller software product. A full rewrite would take the best part of this decade. A more realistic strategy would be to gradually replace the worst parts of the system. Identify the problematic components and fix or redesign them. Sure the UI sucks but it sort of works. The 2.0 development even made it look semi native on some platforms. Deoupling the components of the suit means that you can evolve them (or kill them off) seperately.
No they'll simply ignore him. Has worked just fine so far. It's not like that wasn't very obvious anyway.
Actually you are quite right. The big problem in the linux community is that they are not really open to this kind of criticism. People have been saying for years that UI sucks and for years it has continued to suck.
...).
Lets face it, the GIMP UI is pretty bad. It is going to take some major rearchitecting to fix that. One reason why that won't happen is that the people who are supposed to do that obviously don't get it (just look at the current UI
I have a canon powershot A40, which is now five years old. It's a 2 megapixel camera and was a pretty good thing for the time. Over the years I've learned to make pictures with it that I can fix in photoshop. The ccd sensor provides plenty of detail but the camera software will seriously affect image quality unless you intervene. Images tend to be overexposed, slightly blueish with too much contrast. Now underexposure is really easy to fix. Overexposure on the other hand means white areas in the picture with no detail whatsoever. Similarly, contrast improving algorithms are lossy: they actually throw away a lot of detail. So what I do is shoot pictures with the manual mode that are a bit underexposed. Then I load them up in photoshop and fix them. First I fix the levels. Then, if necessary, I add a curves layer to boost shadow detail a bit. Finally I adjust the color balance to compensate for the blue (especially shots with snow in them). I try not to use the brightness and contrast controls as both are lossy (just look at what happens to the histogram if you use them).
Good camera software should not get in the way of this process. Unfortunately, many cameras do get in the way deliberately. Basically all the low end cameras try to 'fix' the photo in the camera so the users don't have to. The end result with the better ones is not that bad but you'd be able to get much more out of the photo if you'd have the raw material in photoshop. That's why all the highend cameras support the raw format.
What I would like is a compact high end camera. Even the pros adit it: they shoot most shots with some cheap compact camera instead of the bulky SLR they leave home most of the time. I don't have an SLR and I don't think I want one. I do want some of the features though. My compact high end camera of choice would sport great image quality, raw format support, full control over everything (and some convenient defaults for most) and good camera software that does not throw away detail on my behalf.
I'm seriously considering the canon S80. If it had raw support I would have bought it already.
Particularly because running the two on the same platform is a bit hard. Mono runs on windows, sort of, if you want to. But forget about getting any of the mono tools working on windows. And lets be fair, most people currently developing on visual studio are unlikely to run anything else than windows. So even if they wanted to, there's no way for them to even try out monodevelop short of installing linux.
I have to admit Sun seems to have some positive swing lately. They're selling some hardware again. Their open source policy finally seems to be bringing in new customers, after years of shooting themselves in the foot with weird licenses, inconsistent marketing and plain corporate stupidity affecting all their business units.
Their strategy of the past years has been very ineffective. Java has become a multi billion industry, Sun has invested hugely in it but they have failed to cash in on it directly (I suspect they barely break even). Also it hasn't driven hardware sales that much. Their main competitor on the other hand seems to have a very succesful Java strategy. IBM is leading the way in application servers, IDEs (eclipse, rational rose) and middleware with basically the whole industry eating out of their hand, including most JCP specifications committees. On top of that they also sell the hardware & support to go with the software. IBM loves Java!
Then the whole x86 solaris thing has come a long way too (from 'hey it's free now but we support it for a fee', 'oh wait we don't do that anymore' to 'oh well lets open source the whole thing and forget about it' to finally 'hey x86 solaris is really important to us'). I mean, what do they want?
Sparc sales have been a disaster for the past years with people basically favoring IBM power and x86 with linux. They may be laughing at intel for itanium but they have one thing in common with intel: x86 is driving sales for both of them. Sun has a few next generation architectures on the shelves which no doubt they are going to try to sell. These chips had better be way better than the competition (and their mediocre current offerings) or otherwise whatever spin they put on it won't work. Personally I'm not convinced yet. Their two operating systems (solaris/linux) & two architectures (x86/sparc4 aka niagra) policy is going to continue to confuse people. Sun seems to think the combinations can coexist without affecting each others marketshare. I don't. Confused customers will look to IBM and others.
To me Sun still is a company in trouble. Maybe a few of their business units are recovering (finally) but that still leaves large parts of the company not performing very well.
Gates is a bit of an oddball. I can't think of anything particularly brilliant he's ever done and yet he's the richest guy on the planet so he must be doing something right. Arguably making a deal was with IBM was the smartest thing he's ever done (and the stupidest thing IBM has ever done). But from a technology perspective, nothing exciting happened at the time. I mean DOS was a shitty system, even at the time it was introduced.
... well need I remind anyone of his 1995 visionairy masterpiece in which he managed to almost totally ignore the internet? To this very day the guy rarely says or does anything interesting. He's sitting on a huge pile of cash but other than giving it away to charity (which is good) he's not doing anything interesting with it. Somehow, I think Steve Jobs would never be able to just sit on a pile of cash like that. He'd be itching to spend it on something, anything.
Jobs on the other hand has always associated himself with cool stuff though none of it can really be attributed to him. He was just sort of there at the right moment, surrounded by brilliant people doing really great stuff.
If I'd have to pick one it would be Jobs. Mainly because I like people capable of thinking out of the box. There's too few of those in this world. And Jobs has certainly proved that he's capable of that. Gates on the other hand
XHTML is a good illustration why it is a good rule to never finalize a standard without a usable reference implementation. Right now it's just a boring document. I suspect the people who are supposed to implement it have barely read it. And frankly, I don't think they will.
XHTML 2.0 is part of a fictional roadmap that was tossed around late last century and then was discarded. Microsoft decided to sort of stop developing their rendering engine and Netscape was sold to AOL and eventually was liquidated. Microsoft is still pretty happy to not do any significant new standards support in their upcoming browser. Meanwhile the mozilla people, opera people and konqueror/safari people are working outside the w3c to get some real standards work done: like this decade! Html 5.0 is something I can see happen, xhtml 2.0 was dead. on arrival (did they ever bother to make it a recommended spec?). As a matter of fact, XHTML 1.1 is as unlikely to ever be a relevant standard as well. And that one would be relatively easy to implement for browser developers.