The following is a list of items that we think are strong candidates for removal from illumos. We don't think these should be very controversial.
|pcmcia & cardbus, cardservices APIs||None (USB)||There are no pcmcia leaf drivers. There were only ever a tiny handful of useful Cardbus (32-bit PC-Card) devices, pretty much all of them were 100 Mbps NICs. Polls of the community reveal no current users, as everyone is using USB these days. Further, these APIs have a dependency upon the legacy power management code, so removal will facilitate some of that cleanup.|
|Sun E10K, SF12K and SF15K support||None||These platforms depend upon closed code, and are long-obsolete enormous data center class systems. There are no current illumos users, and its inconceivable that there will ever be any in the future. These platforms are far beyond the means of any normal hobbyist. They also have dependencies upon closed-source modules, and have platform details that are undocumented, so its not realistic for us to support them with purely open software.|
|cvcd||None||Network console support for E10K and SF15K, undocumented protocol, the client for which only runs in a service process (which we lack code for).|
|idn||None||Inter-domain networking used on the E10K and SF15K.|
|printmgr||CUPS||Java based GUI for printer administration; most people use CUPS these days.|
|wanboot||inetboot||Support for booting (and hosting booting of) SPARC servers using an HTTP/HTTPs download of their kernel image and root disk. Primary purpose thereof is to support booting across the Internet. No illumos consumers identified, or anticipated (after polling the net). Brings in a private copy of OpenSSL with local customizations, so removal of this facilitates removal of OpenSSL from the gate. (There are other consumers which need to be updated first, though.)|
|pkgadd net installation||curl + local installation||No current illumos consumers for this exists, it uses bits of wan boot infrastructure above, and adds to the OpenSSL problem. (NOTE: SchilliX has a local copy of this and does use net installation, but they maintain a local fork, and no longer are an illumos downstream.)|
|wireless USB||None (bluetooth)||Wireless USB was an alternative to bluetooth. No illumos users exist, very very few devices were ever produced or manufactured, and the industry ceased all efforts to support WUSB.|
|llc2||None (llc1?)||Closed source module, with no active known consumers.|
|warlock||None||Dependencies upon Sun internal tools (compiler support!) that were never made public. Warlock support in illumos has badly bit rotten.|
|legacy power mgmt DDI||None||ddi_dev_is_needed, pm_raise_power, pm_create_components, pm_set_normal_power, pm_get_normal_power, pm_destroy_components; these things are legacy power management APIs, cleaning them up will facilitate other code simplifications.|
Here are some things we think would also be useful to remove, but we would like feedback from the community:
|SUNW,UltraSPARC-IIi-Netract, SUNW,Netra-CP2300, NetraCT family||None||Is anyone using any of these Netra platforms? We don't think so. (Some of them are telco specific, e.g. 220V power.) This includes Netra ct400, ct800, etc., and all Netra CompactPCI platforms.|
|hxge||None||10Gbps ethernet found on blade servers.|
|Sun E3000-Sun E6500 (SUNW,Ultra-Enterprise)||None||These legacy mid-range Enterprise servers were expensive. They have large footprints, and we think its likely nobody is using these with illumos. The support depends upon a closed-source platmod, so this platform can't run without closed code.|
|Sun Fire 3800-6800||None||Serengetti systems, and newer, but still probably too big for most hobbyist use. (8.5 up to 28 RU). These systems EOLd about a decade ago.|
|Sun Fire E3900-6900||None|
Olympus this uses Fujitsu CPUs (I think), these systems are beyond the reach of the usual members of the illumos community.
Every known BSD distribution (FreeBSD, NetBSD, OpenBSD, Darwin/Mac OS X, etc) as well as every known Linux distribution uses the same file(1) implementation (http://www.darwinsys.com/file/). We should just ditch ours and let the distros ship the above. The output isn't identical, that's why this isn't in the "no-brainer" table above.