Interesting, but I don’t understand how Amazon helps you. If you release your software in a form that users can easily upload and run on EC2, presumably you’ll have to simplify the dependencies anyway (Sybase, huh), so users can just as easily run it at home. Or am I missing some part of the story?
No, you can package it as EC2 image—which has identical hardware (virtual), identical software (whatever you put there) etc.
People just clone your image, and start it. It feels very much like starting a new lifeform from genetic instructions from pristine zero state vs trying to genetically engineer existing live organism (a much more difficult thing).
This manner of “releasing” software counts as evil in my book. It certainly goes against the spirit of open source. It’s just like releasing code that depends on an expensive Oracle license.
I don’t see anything evil about it. Are Knoppix-based live cds evil? They’re very similar things.
It’s usually just a standard image of two year old CentOS (or whatever they decided to use—for some reason CentOS seems most popular, even though I’d much rather have Ubuntu everywhere) + a bunch of tarballs + some initialization scripts which unpack and sets it all up.
Very little about EC2 is really EC2-specific (it’s a different matter about S3 and some other Amazon services), so if you have a spare old x86 Linux server and are willing to suffer from exactly the same distribution they used, you’ll be able to get it running in little time even without terribly much clue.
If you want to use different architecture / distribution / or even much more recent CentOS than the image—well, that will require some clue and some effort—but it would anyway, and at least this way people releasing the software don’t have to support it.
The idea of write-once-run-everywhere software is extremely far removed from reality. Even write-once-run-on-every-x86-Linux just doesn’t work in practice. The cause of problems are distros which never got together and set up a single standard, not people writing software that just works on a single distro.
If we run our software here using the same front end, we don’t have to modify it to make a release. That is, we wouldn’t write it to assume use of the Sun grid engine and Sybase if we were using the cloud. (I hope.)
Yes, if we said, “Here’s our software, which will run on the cloud provided you have a Sybase license,” that wouldn’t be a huge improvement.
Also, users usually can’t run these programs at home, since they require dozens or hundreds of computers.
Interesting, but I don’t understand how Amazon helps you. If you release your software in a form that users can easily upload and run on EC2, presumably you’ll have to simplify the dependencies anyway (Sybase, huh), so users can just as easily run it at home. Or am I missing some part of the story?
No, you can package it as EC2 image—which has identical hardware (virtual), identical software (whatever you put there) etc.
People just clone your image, and start it. It feels very much like starting a new lifeform from genetic instructions from pristine zero state vs trying to genetically engineer existing live organism (a much more difficult thing).
Thanks!
This manner of “releasing” software counts as evil in my book. It certainly goes against the spirit of open source. It’s just like releasing code that depends on an expensive Oracle license.
I don’t see anything evil about it. Are Knoppix-based live cds evil? They’re very similar things.
It’s usually just a standard image of two year old CentOS (or whatever they decided to use—for some reason CentOS seems most popular, even though I’d much rather have Ubuntu everywhere) + a bunch of tarballs + some initialization scripts which unpack and sets it all up.
Very little about EC2 is really EC2-specific (it’s a different matter about S3 and some other Amazon services), so if you have a spare old x86 Linux server and are willing to suffer from exactly the same distribution they used, you’ll be able to get it running in little time even without terribly much clue.
If you want to use different architecture / distribution / or even much more recent CentOS than the image—well, that will require some clue and some effort—but it would anyway, and at least this way people releasing the software don’t have to support it.
The idea of write-once-run-everywhere software is extremely far removed from reality. Even write-once-run-on-every-x86-Linux just doesn’t work in practice. The cause of problems are distros which never got together and set up a single standard, not people writing software that just works on a single distro.
If we run our software here using the same front end, we don’t have to modify it to make a release. That is, we wouldn’t write it to assume use of the Sun grid engine and Sybase if we were using the cloud. (I hope.)
Yes, if we said, “Here’s our software, which will run on the cloud provided you have a Sybase license,” that wouldn’t be a huge improvement.
Also, users usually can’t run these programs at home, since they require dozens or hundreds of computers.