3 ms·

Yeah if anything what was too late was solaris' open sourcing. It could have had Linux' lunch. It could have done wish some Linux dev love too, Linux was much

Yeah if anything what was too late was solaris' open sourcing. It could have had Linux' lunch. It could have done wish some Linux dev love too, Linux was much further ahead in terms of updates. I remember updating one of our Solaris servers back in the 9 or 10 days and it was basically just a list of things I had to change manually. Almost nothing automated.

jjav 5y ago
> Yeah if anything what was too late was solaris' open sourcing. It could have had Linux' lunch. I know some disagree but I maintain that the key factor to Sun decline was the unbundling of the compiler when Solaris came out (vs. SunOS where it was included). Stepping back to early 90s, open source was generally never packaged and prebuilt. You had to download the tarball and compile it locally. Installing a new machine meant spending a good while building all the tooling to make it useful. Approximately all of academia and research labs were SunOS shops. All open source code built out of the box (tarball) with ease on SunOS since its developers were almost certainly on SunOS themselves, or at least used it as a primary testing platform given the popularity. In the early days of Linux it was a lot of hard work to get open source to compile, since the code mostly assumed SunOS. As late as the late 90s, it was still easier to compile many things on SunOS than it was on Linux. But with Solaris in the second half of the 90s[1] the compiler was unbundled. Suddenly it was impossible to build all the code you needed to make the system useful, without jumping through a buch of hoops. Either buy an expensive SunCC or go get and install the GNU compiler but then the makefiles tended to assume that if the OS was sunos the compiler was SunCC so you had to go resolve differences between Sun and GNU compilers. Meanwhile, compiling packages on Linux was steadily becoming easier. [1] Well, Solaris came out in the early 90s but was pretty bad compared to SunOS initially. So most SunOS shops, like the research lab where I was, resisted migrating for many years. Solaris 2.5.1 in 1996 was generally considered the first version worth upgrading to from SunOS.

jjav 5y ago

> Yeah if anything what was too late was solaris' open sourcing. It could have had Linux' lunch. I know some disagree but I maintain that the key factor to Sun decline was the unbundling of the compiler when Solaris came out (vs. SunOS where it was included). Stepping back to early 90s, open source was generally never packaged and prebuilt. You had to download the tarball and compile it locally. Installing a new machine meant spending a good while building all the tooling to make it useful. Approximately all of academia and research labs were SunOS shops. All open source code built out of the box (tarball) with ease on SunOS since its developers were almost certainly on SunOS themselves, or at least used it as a primary testing platform given the popularity. In the early days of Linux it was a lot of hard work to get open source to compile, since the code mostly assumed SunOS. As late as the late 90s, it was still easier to compile many things on SunOS than it was on Linux. But with Solaris in the second half of the 90s[1] the compiler was unbundled. Suddenly it was impossible to build all the code you needed to make the system useful, without jumping through a buch of hoops. Either buy an expensive SunCC or go get and install the GNU compiler but then the makefiles tended to assume that if the OS was sunos the compiler was SunCC so you had to go resolve differences between Sun and GNU compilers. Meanwhile, compiling packages on Linux was steadily becoming easier. [1] Well, Solaris came out in the early 90s but was pretty bad compared to SunOS initially. So most SunOS shops, like the research lab where I was, resisted migrating for many years. Solaris 2.5.1 in 1996 was generally considered the first version worth upgrading to from SunOS.