On Sun, May 23, 2010 at 08:42:23AM +0200, Josh Hurst wrote: > On Sun, May 23, 2010 at 6:51 AM, Alan DuBoff <[email protected]> wrote: > >> The comment about GNU is IMO unjustified. The ksh93-integration and > >> AT&T team have done a much better technical job than GNU in the last > >> four years. We got ksh93, a lot of modernized tools, even more in the > >> work, with GNU *and* BSD extensions, stick to POSIX and a stable API > >> and they are evolving with the rest of the open source world. > > > > That all sounds good, but I have long used GNU extensions on Solaris, > > "gtar" as a case in point so that I could get compression support with tar. >
In real life, I'm a sysadmin. Mostly dealing with Linux, also some Solaris. Prior to that, I was supporting technical users working on more than one platform. I had the pain of building GNU tools under Solaris 8 - because people wanted a tar that would work consistently across OS variants. Likewise GCC. Korn shell compatibility was forced on me - I argued a few years ago that we should just put bash everywhere :( > You picked a bad example. GNU tar has its own share of problems. By > default no other achiever than GNU tar can unpack long path names in > archives created by GNU tar. That's a big problem. It becomes worse > because tar archives with long path names created with GNU tar from > 2002 can't be read back by GNU tar from 2010. That's a huge problem. > Of course it can be solved by using the old GNU tar, star from schilly > or AT&T AST pax. But it shows how little the GNU community invests in > stable interfaces and interoperability. I have more examples, at least > one for each GNU tool I know. > > > gupdatedb/glocate are another example. There are many Sun/Solaris folks > > that will be quick to tell you how crappy the GNU extensions are, > > but there are more than a few like me that would just like to have some of > > the features they offer. The bigger problem is in having both code bases. > > In Nexenta's case they don't maintain much of that, they use the GNU base > > and Debian folks maintain that for them. > I think that current Nexenta may be based off Ubuntu - but may be moving back to Debian eventually. The problems lie in CDDL and GNU licence incompatibilities. Both are free software licences accordiing to the OSI but are not mutually compatible. > Maintain may be the wrong word. The quality is substandard (at best). > Look at the source code, but only if you have a healthy heart. Look at > the package configurations, either compile without optimizer or use > -O5000 to ensure gcc bugs really break the application. Look at the > discussions where the maintainers spend more time discussing licenses > than fixing bugs. Look at their obsession with x86, many of the SPARC > packages are in a state of disrepair. Look at GNU tools on SPARC - > many of them work poorly - even the latest coreutils package gives you > SIGBUS because the basic SPARC rule of natural data type alignment is > ignored. These things have been improved in the last years with the > rise of tools like valgrind and new major platforms like ARM. But it's > still a long way to go. Once upon a time, I put a current Debian onto a dual CPU Sparc 20 to go to a national Linux expo. There, alongside Intel boxes, was the self same software running well on Sparc - all of it. Sun folk manning a stand close by couldn't believe it. Debian do care about licences - too much, sometimes. Not everyone had Sparcs to build software on: too much was predicated on closed compilers: too much was difficult to manage because of proprietary adapters. That got better. Sun made strategic mistakes in killing Solaris on x86, then reviving it, then producing fantastic x86 servers which no one was prepared to pay to run Solaris on. Transitive selling to IBM then removed the Sparc - x86 bridge: you can't readily migrate legacy Sparc code to anything else now. Then Sun sold to Oracle - and all bets are now off: even hardware support contracts - and no hobbyist is likely to pay for a support contract on Oracle hardware in order to run Oracle Solaris for personal use. > > > However, not to digress, my point was more in relation to leveraging open > > source to solve the same problems they were designed for. > > To be able to work with the other open source communities so that all can > > grow as a whole. > > +1 > > But Oracle (includes ex-Sun) still has to learn to cooperate and > interact with the/a community. Unfortunately I really have doubt if > they are willing do to just that because they are not able to maintain > a stable and equal partnership, even with established communities and > even after FOUR years. > http://www.mail-archive.com/[email protected]/msg00208.html > is the hallmark of this behavior - Oracle just picks the parts which > fits into their view of the world and the community is locked out if > they do not agree. Of course the community followed ALL of Oracle's > rules like filing PSARC cases and code review. Not enough for Oracle. > IMO an equal and trustworthy partnership looks different. A lot > different. > There isn't a partneership. OGB can't leverage the work on OpenSolaris unless Oracle say so. OpenSolaris, in some sense, was always too little too late and too much was retained as built on a proprietary infrastructure that cannot readily be duplicated. OpenSolaris wasn't (? still isn't) even self-hosting or able to self compile without reliance on development releases of (commercial) Solaris that are now no longer available. > > This is the "Not Invented Here" syndrome. > > Right. It's known as "Not Invented At Sun" syndrome. I remember the > heated PSARC discussions when ksh93 was added and all the bricks Sun > put in the way of the team. It's incredibly now much bullshit the team > had to endure just because Sun really wanted to show off all their > fancy cool rules at once. I think this drove off many people > interested in contributing. > > > I used Nexenta as an example only because they were able to put together a > > distribution that did leverage the GNU base. > > But they not only had a distribution together but had ZFS included on the > > root file system long before (Open)Solaris. > > > > More so I believe that John Plocher's point was spot on, because I was > > always 100% supportive of the fact that the <cough> (Open)Solaris > > community should be built out of 100% open code, so the community doesn't > > have to rely on Oracle, and more importantly so that they CAN build it. > > Without the closed bins I don't think the kernel will build anymore. There > > is a LOT of code in closed, and all the HBA controllers have support by > > means of closed bins as I recall... > > Why aren't be going to put money together and HIRE someone. 100 people > pay $100 each month to hire 2-3 people to eradicate the closed sources > in libc. A company could pay $1000 a month and only 90 people have to > pay to $100. Two companies could pay $2000 and only 80 people have to > pay the $100.... > How're you going to bootstrap it in a reasonable time frame? If Oracle close bits of it to you, you can't do _anything_ about it. Solaris 10 / upcoming Solaris 11 are cash cows for Oracle to milk those few businesses which are tied to high end Sparc - perhaps bank trading floors, major government contracts. Otherwise, Oracle look to be selling an integrated solution to run high end databases on - that's it. Everything else that was invented, innovated with or improved by Sun is now completely irrelevant. > > WTF, we can't even use the <cough> (Open)Solaris name. > > Why don't we rename the organisation then? OGB should hold a contest > and community vote and then move all servers to the new name. This > should be easy. Oracle of course will not be happy but it's their > fault. > Check your licences very, very carefully. Check trademarks and copyright. There has been too much confusion between OpenSolaris the codebase / the purported distribution / the trademark(s). At this point, OpenSolaris is a bit murky. It was possibly originally meant that way :) but that's now a two-edged sword. Disentangling is hard - just look at the potential arguments over whether Schilix, Belenix or Nexenta represents the "true" OpenSolaris > >> Just looking at Debian, Ubuntu and Linux only doesn't make Opensolaris > >> better, it gives only a shadow or at best a petty clone but not a top > >> grade enterprise system. > > > > I think Linux is good enough for the enterprise. If I have to sacrifice the > > ability to build the system from open sources, enterprise doesn't mean $#!T > > at the end of the day. > > Live free or die... > > +1 > Ironically, Debian represents a better Linux for the enterprise than many specifically commercial variants. Always Free, containing many things pre-compiled, readily updatable and stable for a fair period. You don't get the fabled backward compatibility that there was for Solaris binaries - but that came at a price: not least the demise of Sun as we knew it. Just my 0.02 £ > Josh Andy > _______________________________________________ > indiana-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/indiana-discuss _______________________________________________ indiana-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/indiana-discuss
