Hi Michael W, I appreciate your kind and sensible words.
You left some numbered questions & comments. I hoped to share a bit. Another distribution? See below (3) Another Name? No... Brand it OI for short. OI is GREAT. (7) Vote on a userland? Democracy is 2 wolves & a sheep voting on dinner. I don't know if that is the right thing to do. Am I the only one who sees what has happened over the past 25 years between Solaris & GNU & Linux? (4) userland rant coming, take out your knives and guns! GNU userland is basically Linux, because GNU never finished a real kernel, and Linux never had a real UserLand. It was 2 half successful projects merged together, a marriage of convenience but never a true union. SVR4 userland was only ever partially implemented in Solaris, two successful products, each killing off each others most successful features over time. I think there just needs to be an effort to make SVR4 userland what it used to be in other defunct SVRx worlds - easy to use: text, X, and then extended to HTTP interface. GNU can come along for the ride! I still can't believe that FMLI was killed when other SVR4 systems were able to do all system admin work, including: user admin, group admin, printer admin, packaging, network configuration, os upgrade, running user applications, and provide extensibility for data centers to build their own menu based automations. SVRx leveraged the same FMLI language to deliver everything over X (XFMLI) with no code changes. It should have been wrapped in HTML/JavaScript instead of killed (and not replaced with something functionally equivalent.) As soon as Sun had something good, they killed it for a worse user experience. Sunview was nice and fast - it made them a market leader with easy administration. DisplayPostscript brought desktop publishing, then went the way of the dinosaur. OpenWindows was slower, but not ad slow ad what was to come. FMLI arrived, could build menuing interfaces for user and admin communities. OpenStep was beautiful, but never delivered. Motif was a pig when CDE wad introduced. The admintool became more crippled. At least CDE came with windowing sh for building GUI/scripting interfaces, even though xfmli was never ported. Gnome was bigger/slower yet with virtually no simple way to administer/customize it. FMLI and windowing shells got killed with nothing to replace them. User communities, what were they? Once again, more difficult to manage at every step, rip out what worked well, and leave nothing to replace it. Workstations were a breeze to set up. Pull a box off the loading dock, copy down the MAC, copy it into a config file, roll the workstation onto a desk, boot it up. Zones and workstations should have been the same abstraction instead of killing diskless workstations and later killing sparse zones. The workstation handling made Sun market leader, then they stopped advancing state-of-the-art and killed it. Svr4 packaging was awesome. Packages were quick to build via a script. could be verified, packages already installed could be verified in case of tampering, it could be network based with nfs, even worked with zones automatically (made my life easier with a dozen zones per host machine!) Packages could be made architecture an os aware (my packages worked across SPARC Solaris and Intel non-Solaris OS's seamlessly, no need to have multiple packages.) Was it perfect? No, but could have been extended via http with little effort, buy it was killed. What about sar? It was awesome. Someone decided to make the "-u" options basically useless recently by making a column show bogus data. I used to pop graphs out in a one-liner using xterm with tek emulation with sag. Oh, someone decided to kill fast out of the box performance graphing and not replace it. That was ok because "sar -u" gave a bogus column anyway, right? If we are going to get another distribution, how about one based upon Illumos that works on SPARC III, IIIi, IV, IV+ out of the box? If not, don't bother. If we are getting another distribution, how about one that works in 128mb and can be used with embedded systems with standard text menus over a terminal over X console via ssh? Throw in a light and easily customizable window manager like OLVWM? How about make vnc work out of the box with the start of a service? We want a tight distribution to do things with. If not, don't bother making another distribution. Some claim they want a server OS? Do it right. Make the TK SNMP MIB browser shipped with the OS actually work (what a concept) and monitor local & remote systems, use RMON & DISMAN for all NET-SNMP alerts and data gathering, use SNMP TrapD to syslog out-of-the-box for system health, snmp-to-dtrace interface, sar-to-snmp interface, zpool-to-snmp interface, etc.) Make a real server OS with managability light years ahead using incremental improvement instead of making a Linux look-alike (which feels 10 years older with missing features, that are being killed off of Solaris.) Don't claim to want to be a server OS without decent SNMP, if DTrace is really that good, gateway it, and make sure the TK based X MIB browser can graph it all - otherwise, don't bother with another server distribution. If we want a real user workstation distribution, give an option like olvwm so hundreds of users can easily be administered centrally and hundreds of users can painlessly be running virtual desktops on a single socket. Make network boots that look like sparse zones. Provide talk, wall, email, nntp news, smtp, finger, ruptime on the servers and remote workstations (that should boot off the server and look just like a zone.) Throw in blogs & wiki for fun. Ensure there are X and Http interfaces for everything. Other WM's should install through package management and work without configuration. Web browser should auto-update. PDF and Flash should auto-update. Add a multi-platform IM tool. If not, don't bother with another desktop distribution. If a feature was fast & easy to use, it was killed. I am tired of seeing people kill a good thing, reinventing the wheel, just to give the community less, a little slower, but looking more modern. If we want really modern user interface, resurrect Looking Glass. How about OpenStep, to draw some Apple programmers working on top of Darwin? One of the OpenMotif libraries, for compatibility? How about adding OpenLook back in there, for a fast/clean window manager? I know, Unicode is a problem, but not in the Open Source world - deploy what we have, let the community choose, give the community time to fix it. Invest time into OpenIndiana and put back some of the old dead stuff that made SVR4 userland superior to GNU userland... Then take standard SVR4 userland stuff and extend it, that code is static and we have no fear of merging it back. Keep the GNU userland that comes from Solaris, let Oracle upgrade it. If the don't release the code, we have options, but actively developed code that overlaps with Solaris should be the last things we touch. There are several high priority places to target: - SPARC, for disaffected development users and cheap equipment which will be rolling out of data centers, as new equipment is rolled in, OI has an opportunity to make a foothold into these datacenters, as long as old equipment is there and we retain compatibility. - User Friendly Administration, to bring in those new users. Linux did it with a vastly less friendly system from original SunView or SVR3/4 (olwm, cde, fmli/xfmli.) - User Friendly GUI, to make it easy to use & customize, not fancy, consider old hardware off eBay, must do basics out-of-the-box like an Apple, as easy to maintain as an Apple. Without Jobs, this arena should be targeted, this is the future. If it can not be managed easily, it should be an add-on, not default. Users should be able to adjust their monitor settings, right? (I do some of that with vnc terminal servers and scripts, today.) - Embedded and appliance, gotta be easier to manage out of the box than other systems (snmp, telnet, ssh. X, http)... Build every interface as an abstraction so no work must be duplicated. This is our bread-and-butter now. This pays the bills. We should be BETTER than EVERYONE in appliances if we want to survive. OI should become the choice OS for embedded appliances. - Cloud, this is medium future. Joyent is doing awesome. Desktop virtualization should be done like no one is doing today, to make a splash in the trade rags. What about booting OI off a cloud and resurrecting cachefs, make the desktop look like a zone? Leverage free services to host pieces? Use a USB stick to make a dataless workstation? Checksums with zfs should make an awesome back end store with very little front end effort. Zfs compression should make an awesome mechanism to compress data between the client & the cloud. OI should become the choice OS for clouds and virtualized desktops, based on our history. - Virtualization, host other architectures in zones, don't have to do it fast, just do it. Intel, SPARC, ARM, POWER. Base it on KVM. A zone should be able to host any architecture OI instance on any platform. Software should work everywhere, speed should be the only variable. Old Sun, & Mac's should be our friends. Old linksys should be our target. Someday, ARM will be in the cloud, OI should be their first choice - let then develop in a zone, then let them run Intel in a zone. - clustered file system, this Is killing Solaris. I needed one for years. At the very least, we need some kind of file replication. Will Lustre ever be available for OI? Will our changes be too much to make it incompatible when it arrives? We have to build on what we have done well (we have the chance to be the new SVRx), stop scrapping stuff which works (give ISV's another place to make their goods available, before the port to Solaris 11 is done), stop replacing with less functional new stuff, stop making new distributions, stop renaming stuff (no one knew what Sun sold because they renamed it every year) - we have to move forward from where we are. Fujitsu should be building SPARC clusters on OI and not Linux!!!!! I lived this back in the 1980's in another real-time community, when I did some assembly kernel work, this is giving me a nightmare just thinking about it. This is not rocket science, much of it can be done in awk or a cron job. Most of it can be directed and libraried to different higher level contributors (not kernel coders.) I am willing to help. I hope this gives people a better feel for where some of us come from. We just want our 25 year old scripts, books with notes, subroutine libraries, and home-made tools to work. We want more features - not just rip & [almost] replace approximately every 6 years. Dave http://netmgt.blogspot.com/ PS the (5) "Crown Jewels" of Solaris are of little value when we don't see the benefits through standard management interfaces (SNMP for zfs, DTrace, zones, crossbow, netfilter, etc.) and standard user interfaces (syslog viewer, TK snmp grapher, TK snmp walker, user programmable Alerts from syslog, users in a zone "talking" to another zone, etc.) OI should be able to be configured through SNMP as well as configure others OI systems through SNMP. A true Internet Cloud OS as well as Desktop. Sent from my iPhone On Nov 28, 2011, at 2:16 AM, Michael Widmann <[email protected]> wrote: > Hi > > Well spoken everyone - one is missing that doesn't declare himself and should > please said some thing on this : Garret .... > > 1.) Garret in front of the curtain please > > 2.) Would it be frank to ask - could we donate for OpenIndiana? and if so > where - and what do donate? > > 3.) Question: Is there a real problem with the name - or only this IPS / SRV4 > Package hating generation conflict? > > 4.) UserLand discussions are slightly boring (IMHO) - cause the OpenSolaris / > OpenIndiana Userland I'm personally used too not really interested in debian > UserLand or anything else... > > 5.) What really matters: ZFS / DTrace / KVM / Zones / Crossbow - could we > all work together to make a progress their - and maybe starting to innovate > with an "open board" > > 6.) Every Distro has it's beautiful side - could we hammer out (for people > not knowing one of this either) which one is best for what case? (making a > list together where each is aimed to be installed / used) > > 7.) Let the community vote for the userland and the winner should help the > others to integrate .... (if it is illumnos / debian userland - please help > OI to integrate) > > 8.) thanks to everyone who does a great job - on the core / the distribution > / the integration of new things (nexenta - illumnos / alasdair - openindiana > / joyent team - kvm / dtrace and tons of updates and fixes) > > Michael > > > 2011/11/28 Alexander <[email protected]> > Well said. Just let's work together, I do not understand why the name of the > distribution can be an obstacle. I do not see any threat to OI, moreover, I > think working together on the integration of new packages and the use of one > illumos-userland will helps everyone. Let's just work, as Bryan said. > > Sent from my iPad > > On Nov 28, 2011, at 1:43 AM, Bryan Cantrill <[email protected]> wrote: > >> >> On Sun, Nov 27, 2011 at 12:40 PM, Michael Widmann >> <[email protected]> wrote: >> Any comments on this joyents and nextentas? >> >> First, Joyent, Delphix, Nexenta and every other member of the illumos >> community contributes to illumos -- to the core operating system -- which in >> turn benefits everyone (OpenIndiana included). So we have in fact helped >> OpenIndiana (most significantly with our KVM port to illumos, which >> OpenIndiana included in its oi_151a release) -- and we will continue to do >> so. >> >> That said, I think it's important that we as a community recognize that what >> binds us is core OS technologies (ZFS, Zones, Crossbow, DTrace, KVM, etc.), >> and not how those technologies are packaged and distributed. A central >> aspect of the failing of OpenSolaris (in my opinion) was that we >> collectively (and Sun in particular) insisted on there being only One True >> Path for the entire system. At its best, this ethos manifested itself as >> endless discussions on governance and voting and constitutions -- and at its >> worse led to arguments, discord, politicking and fracture. >> >> But with illumos, we have a rebirth: we have not only fresh blood in terms >> of technologists, but also (I would like to think) more tolerance around >> those elements that are ancillary to those core technologies. As such, >> several distributions have flowered that would have not been possible in the >> shadow of OpenSolaris -- and I expect more to come. This is _healthy_ as it >> means that more people (not fewer) will be exposed to our core values as new >> distributions arise to fill new niches. As a community moving forward, we >> need to stay focussed on the values that bind us -- and that means leading >> with the technology, not pre-announcements or rhetoric or endless >> discussion. To that end, I would point to the illumos hackathon as a >> shining example of what we can and should be doing: similarly minded people >> coming together to advance the state of the art in operating systems! >> >> With that, I would like to ask that we cease the friendly fire and get back >> to work. Speaking personally, I am going to be spending the afternoon >> finishing up the ::scalehrtime dcmd that we found so invaluable on a nasty >> KVM problem this past week (patch to come on that one), and adding some code >> to the panic path that would make a similar problem slightly easier to debug >> -- work that I believe to be examples (if extraordinarily small ones) of the >> values that bind our community... >> >> - Bryan >> >> illumos-discuss | Archives | Modify Your Subscription > illumos-discuss | Archives | Modify Your Subscription > > > -- > bayoda.com - Professional Online Backup Solutions for Small and Medium Sized > Companies > illumos-discuss | Archives | Modify Your Subscription ------------------------------------------- illumos-discuss Archives: https://www.listbox.com/member/archive/182180/=now RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be Modify Your Subscription: https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4 Powered by Listbox: http://www.listbox.com
