Who has done the real donkey-work in getting accessibility into GNOME?... Sun.
All the boring but necessary work to turn GNOME from a hacker's paradise into a genuinely usable desktop, not just for the typical American geek, but for everyone across the world. There are a lot of people all over the world working on this, but Sun have a team dedicated to this particular aspect, making all the independantly-developed apps usable (not in a vague "yeah, it works" way, but by applying legal standards (I forget the details, look it up at sun.com) for accessibility.
Not to mention that Sun bought StarOffice when it was pretty basic (remember 5.1?) and open-sourced the code, and then improved it dramatically.
Abiword is okay as a WordPad upgrade, but OpenOffice.org has more power than AbiWord.
I get the feeling that the FOSS world is about to repeat what the PC world went through in the 1980s - when Lotus 1-2-3, WordPerfect, Microsoft and other players all battled it out claiming "check-marks" against each other, adding to the bloat of each piece of software, to no real benefit of the end-user. I hope not.
In practice, I suppose the main thing for "market share" is to beat MS Office; for quality, MS Office is no guideline for a sane developer to take. However, that will take serious investment in working out what is actually *needed* by users. It's easier to take pointers from other software. I suspect that any "winner" will be the innovative one, in the end.
BS. The reasons for forking Emacs and GCC were purely technical - nothing to do with licensing.
Sun's reasoning is more practical - they are trying to avoid the mythical examples you cited above.
So - hey - you've just argued for Sun's licensing model. Congratulations.
If I floor the gas in neutral, the engine will seize up
Hope it's under warranty. It should just rev - after all, it's in neutral.
If I start the engine, put it in 5th, and can't get moving, then that's a "didn't work" situation, but then, what kind of idiot would expect it to (given that, AFAIK, every nation requires drivers to pass a test before being allowed to drive a car on a public road)?
Maybe some level of IT Competence should be passed (and enforced) before allowing people to access the internet?
Chase... I just checked that one out.
Yes, I can type in a login / password to the http:/// site, but the form directs to an https:/// website. The login/password is secure, because they are sent to an SSL-encrypted page.
I recently used www.davidlloydleisure.co.uk, which used the opposite (totally insecure) approach: Enter your credit card details into an SSL-encrypted page, only to have that submitted to an unencrypted page!
The other point is valid, though - these firms should use subdomains, not other domains. If I choose to accept that the mybank.com is indeed under the control of My Bank, then I have a decent chance of being able to trust secure.mybank.com, creditcard.mybank.com, debitaccount.mybank.com, etc.
Photos for insurance purposes is a good point... my dad took photos of all the books in the house (they have thousands of books) as the insurance company would be unlikely to believe that they had so many books in the first place... then stashed the photos in the safe... in the master bedroom. Useful, until the house burns down!
The OSI stack is irrelevant IRL, but it is still well worth knowing, and is even useful in understanding TCP/IP.
God knows, enough people who think they understand TCP/IP are woefully missing details on the lower layers; a bit of OSI understanding goes a long way, especially on the IP side of things.
Of course, there's no point saying that here, but I will do... old fart that I am.
But just because a lawyer didn't right it, doesnt mean it wont stand up.
No, but at least the spelling would be correct.
Seriously, get a grip. If the employer wants to use the code for certain things now, they don't know what they'll want to do with it in future; neither does the original poster know. That's why it's worth getting it down in writing before it gets awkward - that "gentleman's agreement" is suddenly worth sh1t once one side of the agreement finds a way to make money from what otherwise sound to be trivial library-style routines.
Let's just have a quick look at http://dev.mysql.com/downloads/mysql/4.1.html for an example- oh yes, there's a Solaris version, an HPUX version, SGI, OSF, lots of different packages for different flavours of UNIX, whereas there's only... oh wait, there are different versions for all the different flavours of Linux, and some of the permutations of glibc. If you've got a different glibc, then sorry, you'll have to build it yourself. If your distro doesn't support RPM (but you do have the right glibc), you're stuck with the tarball.
Linux is better of how?
Of course, it's pointless counting the different architectures (x86, x64, SPARC, S/390, etc) as there's a genuine need for different binaries there, even for one single platform.
If you can view more than a few pages of any one book, and you don't think it falls into one of those categories, you should submit it as a bug.
I may be rather anti-litigious, but in this situation, I think that filing a bug report isn't really the answer. I'm quite sure that legal counsel would confirm that the correct action would be to sue.
Sort of like copying a work in RAM, and/or across network devices is copying?
No, like going to a library, scanning an entire book, taking the scan back to the office, OCR'ing it all, and storing it all in a publically-accessible database is copying. Get the difference?
I can't see anybody having a problem with that; Google respect/robots.txt, and the copyright holders have delegated the authority of passing out that copyrighted material to webserver software. That software is acting on behalf of the (c) holder, and happily giving out copies to any piece of software which uses the HTTP protocol to request the document.
It'd be pretty tough to set up a system of "anyone can have this document" but then say "Hey, you've got a copy of this (which I gave you, sans proviso) - stop it now!" - especially when the protocol has already been configured with a robots.txt specification, which Google (if not all robots) respect.
If you want to follow links from a Google-cached item which doesn't exist, just enter the URL you want to follow into google.com and choose the "cache" option.
Eg - if you want to view Google's cache of http://steve-parker.org/sh/sh.shtml, then just type "http://steve-parker.org/sh/sh.shtml" into Google, and select the "Show Google's cache of..." option.
If you can bot that, then you've got it botted (is that a word?!)
Did you RTFA? Can you find a way of highlighting certain text within a scan without doing OCR?
Of course they OCR the scans. They're not being "low-tech", they're providing images of the books, typesetting, images et. al. An OCR scan might be more useful to google's users, but they're setting out to provide snippets (or entire pages) of actual books. How useful is an OCR scan of "The Very Hungry Caterpillar" without the pictures?!
isn't Games why most people running Linux maintain a Windows partition
No - I recently cut space for a Windows partition (I'm generous with disk space for useful OSes, so I gave WinXP 2.5Gb!) because my employer said that I have to run a particular Windows application. Not a chance of running it under Wine, it's the internal work of [ some hardware manufacturer partnership which (just to eliminate the obvious) doesn't involve English-speaking countries as major contributors ] and it's barely usable under Windows, let along get it going under Wine.
So I said okay then, 1Gb for Windows, 200Mb for the app, another 200Mb for its database, let's splash out and double it to 2.5Gb.
Of course, I had to install Thunderbird and Firefox, and it turned out I also needed MS Excel (the macros barely work in Excel, and we've still not worked out exactly what you need to do to get Excel working "properly" with these macros, it's just guesswork. Our best solution only pops up 2 meaningless "OK" boxes before running the macros. My best summary is that you have to drop your Internet Explorer security levels... suffice it to say that OO.o didn't stand a chance faced with this quality of coding ability).
Being Windows with a mail client and web browser, I had to install antivirus software, too. That partition is currently standing at 98% used with almost no data on it.
What Google is doing is simply saying, "if you care about your book, just let us know."
Okay, I'm going to rip CDs, DVDs and books, then distribute them via the internet. If a particular band, film producer or author doesn't want me to distribute a specific album, film or book, just let me know, and I promise to stop.
Yeah, that would work. The copyright holders will be happy with that compromise.
mp3.com: Rip CD, transmit to 3rd party, 3rd party stores music. What happens next is out of mp3.com's control (though probably still mp3.com's responsibility)
google.com: Scan and OCR a book, transmit to 3rd party, 3rd party stores book (in browser cache, if not explicitly saving to disk).
Not a world of difference, though it's probably fairer to compare with iTunes which also "gives away" 30-second snippets of songs, whilst Google (now) transmits snippets of books. Earlier, they tried a system using cookies to allow a user to read a limited number of pages per day.
Isn't it interesting that (whilst this discussion is mixed about the pros/cons of OO.o) nobody mentions that most of the developers are employed by Sun.
The slashdot trend is to diss Sun, whilst there's about 50/50 support for OOo.
From TFA: "1 RedHat, 80 Sun, and 8 Novell hackers" - but the slashdot ethos still seems predomainanlty anti-Sun.
I don't work with Sun any more (I used to work alongside them), so I've no axe to grind, but it seems amazing that whenever OOo is mentioned, Sun barely gets a look-in, despite the number of users.
The principle of LSB is that you can do it with LSB. TFA points out that you can't. Not only do LSB-Certified distro's (http://www.opengroup.org/lsb/cert/display_product .tpl?CALLER=cert_prodlist.tpl&_pr_id=564 was the example cited by Drepper... it's SuSE for those too lazy to follow a link) not actually pass the certifications under real conditions, I have also seen LSB issues from a simple shell script (http://speedtouchconf.sf.net/). LSB specifies install_initd but doesn't specify its usage. Great... a standard way to install an init script. Totally useless without a standard usage for the script.
I know it's an AC rant, but it contains a valid point.
I constantly have distros compleatly eat themselves trying to do an application update 6+ months after initial install.
BS. So long as the distro is supported, it'll work.
I'd say that Linux position now is far worse than the old UNIX fragmentation and from what I see is constantly getting worse, not better.
Spot on. Maybe too many people here are too young to remember what happened with the UNIX Wars back in the 80/90s, but dealing with PITA distro's (which is most of them) seems worse now than it was when UNIX was bad. Now, I'd say that Linux is getting so bad that UNIX is by default easier to develop for than Linux.
I think that ajs318 has a certain view on computers which isn't necessarily compatible with enterprise datacentre management.
ajs's PoV is fine with home-user and possibly with small-company systems, where a little bit of time spent building a home-grown, unsupported installation is convenient.
Enterprise customers need two things: Support and Compatability.
They need to get the features they need without breaking a support contract (which generally means running binaries provided by the vendor), and they need it to Just Work (TM).
As I mentioned in a previous post, Solaris 2.x is binary compatible from 2.5 to Solaris 10. That means much more than source compataibility to customers... so what if I could compile it from source, I know that I can get a Solaris package and install it on my brand-new Solaris 10 system as well as my decade-old Solaris 2.5 system. Same pacakge, same binaries. That matters a lot to a lot of people (and they tend to be the people who have a lot of money to spend).
Source compatibility is great; Binary compatability is great.
If you can live in a world where you can afford to recompile stuff all the time, then API compatability is what you want; If you can't afford / don't want to rebuild software all the time, then ABI compatability is your thang.
If program X requires program Y version A.B.C, then it requires A.B.C.
Is that so difficult to understand?
Program Y is written by a different team, over whom Program X's developers have no control.
What do you expect?
If GTA3 requires DirectX 9, you'd install it, right?
If the FireFox RPM you downloaded requires GLIBC 2.3 and you've got GLIBC 2.2, then that is your problem, not the developer's problem.
Firstly, that's a binary compatability problem (not seen as an issue in the GNU/Linux world; Solaris is binary compatible back to 2.5, which was released back in 1994 or thereabouts; GNU/Linux focuses on source compatability.)
If Program Y version A.B.C provides certain functionality necessary for Program X, how can Program X possibly work without Program Y version A.B.C?!!!
Gotta love those specially formatted comments! Need to generate or parse this code which is embedded after initial comments and before shell script code? Or need to make a programmatic way of modifying it? Get out sed or perl and re-invent the broken wheel yet again.
What's "broken" about sed or perl? Not that it's relevant, so let's move on...
If you change what the init script does, you're a human being, manually editing a system file. What does it hurt you to update some metadata? Is it easier for some random sysadmin to edit a few simple lines of comments, or to make sure that the XML is fully compliant?
XML is not the greatest thing. But at least it has the ability to replace the current mish-mash of configuration file formats and symbolic links and heavy dependancies which are a nightmare!
XML certainly isn't the greatest thing. It basically requires the entire XML file to be read into RAM to validate and then process it (which is what slows down OO.o, often). XML is just a new mish-mash.
What's wrong with symbolic links? The traditional SysV structure of/etc/init.d/ and/etc/rc[x].d/ containing links was simple and transparent.
Why does every executable need to employ it's own special ascii parsing function?
It doesn't. Init simply runs the scripts. This clobber is added for LSB compatibility (the start of this thread, remember?). LSB isn't implemented well - I maintain a GPL'd shell script (speedtouchconf.sf.net) which installs an init script. LSB specifies install_initd, but doesn't specify its usage, so various distro's implement it however they see fit. The mish-mash of Linux distro's and their different init methods is a royal PITA for me, and for people who use the script.
That's not going to be fixed by XML... any ideas how a shell script can format XML, without reinventing the broken wheel???
Be realistic. If your webapp is a few years old, then it's probably about time that it was rewritten anyway, just to incorporate CSS/XHTML and other stuff which didn't exist / wasn't mature enough when the project started.
Also, if it's a webapp, it's probably a bunch of hacks thrown together anyway - why not take version 1 as a "write one to throw away" prototype, and rewrite the entire thing, using the lessons learned? If you're doing that, there's no pain (and potential gain, if you choose the right language) in switching languages at the same time.
If you're now looking at switching from MS-only to other platforms, it's an ideal opportunity to write platform-independant code which can run against various databases, too.
When you look at the code rewrites to "migrate", it could well turn out easier to rewrite in another language, given the lessons learned in version 1.0. It could well be that you learn that C# is the right language, but it's well worth considering.
After all, you've written so many thousand lines of code, but the work you've done is in creating so many tens or hundreds of algorithms, not the TLOC.
If you now know what features you need to implement, writing them afresh in another language does give you the opportunity to shelve all those nasty API dependancies you've gathered over the past few years.
Note that this doesn't apply so well to mature applications - it can actually be beneficial to maintain 1960s COBOL code versus rewriting it, but a 2-yr-old WebApp is already old, will have grown with its requirements, and is more likely to need a rehaul than a 30-yr-old COBOL app
Slashdot Mirror
Who has done the real donkey-work in getting accessibility into GNOME? ... Sun.
All the boring but necessary work to turn GNOME from a hacker's paradise into a genuinely usable desktop, not just for the typical American geek, but for everyone across the world. There are a lot of people all over the world working on this, but Sun have a team dedicated to this particular aspect, making all the independantly-developed apps usable (not in a vague "yeah, it works" way, but by applying legal standards (I forget the details, look it up at sun.com) for accessibility.
Not to mention that Sun bought StarOffice when it was pretty basic (remember 5.1?) and open-sourced the code, and then improved it dramatically.
Abiword is okay as a WordPad upgrade, but OpenOffice.org has more power than AbiWord.
I get the feeling that the FOSS world is about to repeat what the PC world went through in the 1980s - when Lotus 1-2-3, WordPerfect, Microsoft and other players all battled it out claiming "check-marks" against each other, adding to the bloat of each piece of software, to no real benefit of the end-user. I hope not.
In practice, I suppose the main thing for "market share" is to beat MS Office; for quality, MS Office is no guideline for a sane developer to take. However, that will take serious investment in working out what is actually *needed* by users. It's easier to take pointers from other software. I suspect that any "winner" will be the innovative one, in the end.
BS. The reasons for forking Emacs and GCC were purely technical - nothing to do with licensing. Sun's reasoning is more practical - they are trying to avoid the mythical examples you cited above. So - hey - you've just argued for Sun's licensing model. Congratulations.
When is the spellchecker due?
Hope it's under warranty. It should just rev - after all, it's in neutral.
If I start the engine, put it in 5th, and can't get moving, then that's a "didn't work" situation, but then, what kind of idiot would expect it to (given that, AFAIK, every nation requires drivers to pass a test before being allowed to drive a car on a public road)?
Maybe some level of IT Competence should be passed (and enforced) before allowing people to access the internet?
I recently used www.davidlloydleisure.co.uk, which used the opposite (totally insecure) approach: Enter your credit card details into an SSL-encrypted page, only to have that submitted to an unencrypted page!
The other point is valid, though - these firms should use subdomains, not other domains. If I choose to accept that the mybank.com is indeed under the control of My Bank, then I have a decent chance of being able to trust secure.mybank.com, creditcard.mybank.com, debitaccount.mybank.com, etc.
Photos for insurance purposes is a good point... my dad took photos of all the books in the house (they have thousands of books) as the insurance company would be unlikely to believe that they had so many books in the first place ... then stashed the photos in the safe... in the master bedroom. Useful, until the house burns down!
God knows, enough people who think they understand TCP/IP are woefully missing details on the lower layers; a bit of OSI understanding goes a long way, especially on the IP side of things.
Of course, there's no point saying that here, but I will do... old fart that I am.
No, but at least the spelling would be correct.
Seriously, get a grip. If the employer wants to use the code for certain things now, they don't know what they'll want to do with it in future; neither does the original poster know. That's why it's worth getting it down in writing before it gets awkward - that "gentleman's agreement" is suddenly worth sh1t once one side of the agreement finds a way to make money from what otherwise sound to be trivial library-style routines.
Linux is better of how?
Of course, it's pointless counting the different architectures (x86, x64, SPARC, S/390, etc) as there's a genuine need for different binaries there, even for one single platform.
I may be rather anti-litigious, but in this situation, I think that filing a bug report isn't really the answer. I'm quite sure that legal counsel would confirm that the correct action would be to sue.
Sort of like copying a work in RAM, and/or across network devices is copying?
No, like going to a library, scanning an entire book, taking the scan back to the office, OCR'ing it all, and storing it all in a publically-accessible database is copying. Get the difference?
That's certainly not how I understand US or UK copyright law.
It'd be pretty tough to set up a system of "anyone can have this document" but then say "Hey, you've got a copy of this (which I gave you, sans proviso) - stop it now!" - especially when the protocol has already been configured with a robots.txt specification, which Google (if not all robots) respect.
Google are copying the books without the copyright owner's permission, AFAICT from the information provided. That is breach of copyright.
If you want to follow links from a Google-cached item which doesn't exist, just enter the URL you want to follow into google.com and choose the "cache" option.
..." option.
Eg - if you want to view Google's cache of http://steve-parker.org/sh/sh.shtml, then just type "http://steve-parker.org/sh/sh.shtml" into Google, and select the "Show Google's cache of
If you can bot that, then you've got it botted (is that a word?!)
Did you RTFA? Can you find a way of highlighting certain text within a scan without doing OCR?
Of course they OCR the scans. They're not being "low-tech", they're providing images of the books, typesetting, images et. al. An OCR scan might be more useful to google's users, but they're setting out to provide snippets (or entire pages) of actual books. How useful is an OCR scan of "The Very Hungry Caterpillar" without the pictures?!
No - I recently cut space for a Windows partition (I'm generous with disk space for useful OSes, so I gave WinXP 2.5Gb!) because my employer said that I have to run a particular Windows application. Not a chance of running it under Wine, it's the internal work of [ some hardware manufacturer partnership which (just to eliminate the obvious) doesn't involve English-speaking countries as major contributors ] and it's barely usable under Windows, let along get it going under Wine.
So I said okay then, 1Gb for Windows, 200Mb for the app, another 200Mb for its database, let's splash out and double it to 2.5Gb.
Of course, I had to install Thunderbird and Firefox, and it turned out I also needed MS Excel (the macros barely work in Excel, and we've still not worked out exactly what you need to do to get Excel working "properly" with these macros, it's just guesswork. Our best solution only pops up 2 meaningless "OK" boxes before running the macros. My best summary is that you have to drop your Internet Explorer security levels... suffice it to say that OO.o didn't stand a chance faced with this quality of coding ability).
Being Windows with a mail client and web browser, I had to install antivirus software, too. That partition is currently standing at 98% used with almost no data on it.
Okay, I'm going to rip CDs, DVDs and books, then distribute them via the internet. If a particular band, film producer or author doesn't want me to distribute a specific album, film or book, just let me know, and I promise to stop.
Yeah, that would work. The copyright holders will be happy with that compromise.
mp3.com: Rip CD, transmit to 3rd party, 3rd party stores music. What happens next is out of mp3.com's control (though probably still mp3.com's responsibility)
google.com: Scan and OCR a book, transmit to 3rd party, 3rd party stores book (in browser cache, if not explicitly saving to disk).
Not a world of difference, though it's probably fairer to compare with iTunes which also "gives away" 30-second snippets of songs, whilst Google (now) transmits snippets of books. Earlier, they tried a system using cookies to allow a user to read a limited number of pages per day.
The slashdot trend is to diss Sun, whilst there's about 50/50 support for OOo.
From TFA: "1 RedHat, 80 Sun, and 8 Novell hackers" - but the slashdot ethos still seems predomainanlty anti-Sun.
I don't work with Sun any more (I used to work alongside them), so I've no axe to grind, but it seems amazing that whenever OOo is mentioned, Sun barely gets a look-in, despite the number of users.
Oh well, it's slashdot,
The principle of LSB is that you can do it with LSB. TFA points out that you can't. Not only do LSB-Certified distro's (http://www.opengroup.org/lsb/cert/display_product .tpl?CALLER=cert_prodlist.tpl&_pr_id=564 was the example cited by Drepper... it's SuSE for those too lazy to follow a link) not actually pass the certifications under real conditions, I have also seen LSB issues from a simple shell script (http://speedtouchconf.sf.net/). LSB specifies install_initd but doesn't specify its usage. Great... a standard way to install an init script. Totally useless without a standard usage for the script.
I constantly have distros compleatly eat themselves trying to do an application update 6+ months after initial install.
BS. So long as the distro is supported, it'll work.
I'd say that Linux position now is far worse than the old UNIX fragmentation and from what I see is constantly getting worse, not better.
Spot on. Maybe too many people here are too young to remember what happened with the UNIX Wars back in the 80/90s, but dealing with PITA distro's (which is most of them) seems worse now than it was when UNIX was bad. Now, I'd say that Linux is getting so bad that UNIX is by default easier to develop for than Linux.
ajs's PoV is fine with home-user and possibly with small-company systems, where a little bit of time spent building a home-grown, unsupported installation is convenient.
Enterprise customers need two things: Support and Compatability.
They need to get the features they need without breaking a support contract (which generally means running binaries provided by the vendor), and they need it to Just Work (TM).
As I mentioned in a previous post, Solaris 2.x is binary compatible from 2.5 to Solaris 10. That means much more than source compataibility to customers... so what if I could compile it from source, I know that I can get a Solaris package and install it on my brand-new Solaris 10 system as well as my decade-old Solaris 2.5 system. Same pacakge, same binaries. That matters a lot to a lot of people (and they tend to be the people who have a lot of money to spend).
Source compatibility is great; Binary compatability is great.
If you can live in a world where you can afford to recompile stuff all the time, then API compatability is what you want; If you can't afford / don't want to rebuild software all the time, then ABI compatability is your thang.
Simple, eh?!
Is that so difficult to understand?
Program Y is written by a different team, over whom Program X's developers have no control.
What do you expect?
If GTA3 requires DirectX 9, you'd install it, right? If the FireFox RPM you downloaded requires GLIBC 2.3 and you've got GLIBC 2.2, then that is your problem, not the developer's problem.
Firstly, that's a binary compatability problem (not seen as an issue in the GNU/Linux world; Solaris is binary compatible back to 2.5, which was released back in 1994 or thereabouts; GNU/Linux focuses on source compatability.)
If Program Y version A.B.C provides certain functionality necessary for Program X, how can Program X possibly work without Program Y version A.B.C?!!!
What's "broken" about sed or perl? Not that it's relevant, so let's move on...
If you change what the init script does, you're a human being, manually editing a system file. What does it hurt you to update some metadata? Is it easier for some random sysadmin to edit a few simple lines of comments, or to make sure that the XML is fully compliant?
XML is not the greatest thing. But at least it has the ability to replace the current mish-mash of configuration file formats and symbolic links and heavy dependancies which are a nightmare!
XML certainly isn't the greatest thing. It basically requires the entire XML file to be read into RAM to validate and then process it (which is what slows down OO.o, often). XML is just a new mish-mash.
What's wrong with symbolic links? The traditional SysV structure of /etc/init.d/ and /etc/rc[x].d/ containing links was simple and transparent.
Why does every executable need to employ it's own special ascii parsing function?
It doesn't. Init simply runs the scripts. This clobber is added for LSB compatibility (the start of this thread, remember?). LSB isn't implemented well - I maintain a GPL'd shell script (speedtouchconf.sf.net) which installs an init script. LSB specifies install_initd, but doesn't specify its usage, so various distro's implement it however they see fit. The mish-mash of Linux distro's and their different init methods is a royal PITA for me, and for people who use the script.
That's not going to be fixed by XML... any ideas how a shell script can format XML, without reinventing the broken wheel???
Steve
Be realistic. If your webapp is a few years old, then it's probably about time that it was rewritten anyway, just to incorporate CSS/XHTML and other stuff which didn't exist / wasn't mature enough when the project started. Also, if it's a webapp, it's probably a bunch of hacks thrown together anyway - why not take version 1 as a "write one to throw away" prototype, and rewrite the entire thing, using the lessons learned? If you're doing that, there's no pain (and potential gain, if you choose the right language) in switching languages at the same time. If you're now looking at switching from MS-only to other platforms, it's an ideal opportunity to write platform-independant code which can run against various databases, too. When you look at the code rewrites to "migrate", it could well turn out easier to rewrite in another language, given the lessons learned in version 1.0. It could well be that you learn that C# is the right language, but it's well worth considering. After all, you've written so many thousand lines of code, but the work you've done is in creating so many tens or hundreds of algorithms, not the TLOC. If you now know what features you need to implement, writing them afresh in another language does give you the opportunity to shelve all those nasty API dependancies you've gathered over the past few years. Note that this doesn't apply so well to mature applications - it can actually be beneficial to maintain 1960s COBOL code versus rewriting it, but a 2-yr-old WebApp is already old, will have grown with its requirements, and is more likely to need a rehaul than a 30-yr-old COBOL app