Messages in this thread | | | Date | Sat, 8 Mar 97 20:07 GMT | From | (Alan Cox) | Subject | Linux Networking Status/Plans as of 2.1.28 |
| |
Protocols:
Appletalk: This seems to be basically happy now. There are some bits of localtalk and hopefully some new localtalk drivers to go in over time. Appletalk over PPP I've got no idea about as it needs someone to hack up the PPPD code and also the innards of netatalks user space.
AX.25: Still a lot of work in progress. Most of this is cleaning up things and improving on facilities. DAMA will I hope be back in shortly
Bridge: Doesn't work. Don't intend to fix it in the near future. The right way to do this is to create a virtual bridge device that has a group of actual network devices under it, NOT the way it is done now. There are a big collection of changes and bug fixes to make it 802.1d compliant rather than Digital Lanbridge (bluebook equivalent). Nobody is working on this at the moment.
Decnet: May or may not occur: Remains to be seen
Ethernet: The support code should be pretty much constant now, at least its interfaces.
IPv4: Bugs in ip_fragmentation, some races it would appear from dumps. TCP needs some work doing on it. Routing has some clean ups pending. Alexey is promising serious magic with his next patches so as soon as I get them we find out. There may be some interface breakage to come. Other folk are working on syn-protection/syn-cookies. Nobody is currently working on the TCP problems, but we may well get some of that from the 2.0.29ISS work.
SMBfs over IPv4 apparently doesn't work. I don't think anyone is working on this issue.
IPv6: Probably needs a new maintainer now. No work in progress.
IPX: Seems to have some small bugs. Not having an IPX network I can't tell. The changes affecting IPX are fairly small so it shouldn't take much to fix any problems that may be left. No maintainer, no work in progress. No interface changes anticipated. NCPfs seems to need fixing too.
LAPB: Work in progress but seems to be working nicely.
NETBEUI: Subject to other projects. At the moment paused for lack of time.
NEROM: Should be fairly stable and not have too many pending changes.
ROSE: Coming along nicely. Recent releases have knocked out some of the more obscure bugs.
UNIX: No real "bugs" but some oddities and things that ought to work in better ways. Reports that it is running more slowly than 2.0 but no obvious cause. No work in progress
WANROUTER: Needs some tidy ups, basically works. At the moment only the Sangoma drivers use this interface for WAN products. I'd hope as we get more free wan drivers they use the same inteface. No work in progress.
X25: Work in progress. Seems fairly solid. Someone needs to write some Z8530 sync drivers for a Sparc IPC (or tell me what cabling and Z8530 settings I need to get one talking 9600 baud sync to a camtec X.25 PAD.
Device Layers:
Ethernet interfaces should be stable now. I do want to tidy all the init code up one day, but not until everything is solid. People with ethernet driver bugs should try and fix them on the basis the code interface is stable. All the drivers need support for tx/rx byte counters adding. Realistically for IPv6 and because of things like IPv4 multicasting and appletalk all drivers need to do something sensible with multicasts.
Serial network drivers should also all be a stable interface. The IPv6 PPP code is just a stub, someone needs to write the IPv6 PPP CP into PPPD.
WAN drivers - please make any new ones use the wanrouter interface
Core Code:
The core protocol/socket interface and the low level xx_rcv() interfaces should now be stable. The 'neighbour discovery' interfaces may not be. Several protocols could be improved by using the generic neighbour discovery code. The device/protocol variable stuff may change a lot in ANK's patches - for the cleaner however.
Various "NULL" values will get replaced with "sock_no_xxx" and similar routines for speed and simplicity in some areas.
There are various firewall and security related issues to resolve. One big problem here is that adding the socket security API may cause export problems.
| |