Sun thumbs its nose at a 30 year-old goal (including its own) - SUN

This is a discussion on Sun thumbs its nose at a 30 year-old goal (including its own) - SUN ; Protocol independence has been the goal of network programming since the beginning. But for X.25, Sun blithely pushes its NLI interface - which is basically hard-coded X.25. It shares no abstractions with any other API and is not compatible with ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Sun thumbs its nose at a 30 year-old goal (including its own)

  1. Sun thumbs its nose at a 30 year-old goal (including its own)

    Protocol independence has been the goal of network programming since the
    beginning. But for X.25, Sun blithely pushes its NLI interface - which is
    basically hard-coded X.25. It shares no abstractions with any other API
    and is not compatible with any of the facilities that support all other
    protocols as a group. Plus, it is complicated to use (despite that Sun
    uses happy tones to market it).

    Somebody at Sun ought to take responsibility.

  2. Re: Sun thumbs its nose at a 30 year-old goal (including its own)

    On 2007-04-18 08:48:52 +0100, andrew@cucumber.demon.co.uk (Andrew
    Gabriel) said:

    > X.25 is a protocol which is at the end of its life. It will
    > continue to exist for years, but usage will decline and no one
    > is producing any new drivers for it nowadays. Otherwise it
    > would be worth taking a look at improved APIs, particularly in
    > the light of SCTP which has a number of similar features.


    There's at least one X.25 gateway box that I'm aware of which you can
    connect to over TCP/IP (using RFC 1006), which at least gives you other
    options than using Sun's X.25.

    Cheers,

    Chris


  3. Re: Sun thumbs its nose at a 30 year-old goal (including its own)

    On Wed, 18 Apr 2007 07:48:52 +0000, Andrew Gabriel wrote:

    > In article ,
    > "Charles T. Smith" writes:
    >> Protocol independence has been the goal of network programming since the
    >> beginning. But for X.25, Sun blithely pushes its NLI interface - which
    >> is basically hard-coded X.25. It shares no abstractions with any other
    >> API and is not compatible with any of the facilities that support all
    >> other protocols as a group. Plus, it is complicated to use (despite
    >> that Sun uses happy tones to market it).
    >>
    >> Somebody at Sun ought to take responsibility.

    >
    > X.25 doesn't look enough like many modern protocols to survive very much
    > abstraction, unless you are restricting yourself to a small subset, which
    > most applications running over it don't. X.29 which you referred to in an
    > earlier posting uses all those features of X.25 which nowadays look rather
    > strange in a protocol.
    >
    > An X.25 connection effectively gives you three data streams in each
    > direction, two of which support infinate length records but can't be used
    > together (unqualified and qualified data), and the third which supports
    > only short length records with no flow control (interrupt data), not to
    > mention resets which discard all data and take the connection back to its
    > initial state. Some of these features have appeared in SCTP, but they
    > just don't exist in the classical protocols used on UNIX.
    >
    > One pitfall of NLI is making the packet and window sizes visible to the
    > programmer, and I've seen programmers drive that wrongly and wonder why
    > their application doesn't work (or crawls at a snail's pace). So yes, you
    > do have to know some of the details of the X.25 protocol in order to use
    > NLI.
    >
    > NLI is not of Sun's making -- it is a vendor standard for X.25 support
    > which virtually all UNIX X.25 implementations used. Most of the NLI
    > implementations were derived from the original Spider implementation of
    > the 1980's. (Sun's did also originally, but they rewrote their own
    > implementation at some point, possibly at SunLink 8.)
    >
    > Remember, X.25 predates ISO protocols. Attempts to draw it in the OSI
    > 7-layer model are retrospective refitting, and it doesn't fit too well,
    > although sucessive enhancements to X.25 level 3 over the years was an
    > attempt to make it fit enough of OSI layer 3 so that ISO layer 4 protocols
    > could use it (NSAP addressing, longer interrupt packets).
    >
    > X.25 is a protocol which is at the end of its life. It will continue to
    > exist for years, but usage will decline and no one is producing any new
    > drivers for it nowadays. Otherwise it would be worth taking a look at
    > improved APIs, particularly in the light of SCTP which has a number of
    > similar features.



    I wouldn't mind if the implementation used non-abstract constructs to
    access features of x.25 that aren't in that "small subset" (only because
    it is truly EndOfLife) - but sockets and Sun already offer x.25 facilities
    that are compatible with protocol independence - Sun is throwing the baby
    out with the bathwater.

    I mean, I guess there is grudging support for x.25 sockets and tli, but I
    posed the same question to Sun that I essentially posed here - why does
    Sun recommend NLI - and the answer I got was paragraphs that said because
    Sun recommends NLI.

    I suppose I could accept that NLI was the coup-de-grace for X.25 - if Sun
    would just publish a white paper that would discuss how it compared
    to their socket and tli implementations.


  4. Re: Sun thumbs its nose at a 30 year-old goal (including its own)

    On 2007-04-18, Tim Bradshaw wrote:
    > On Apr 18, 6:12 am, "Charles T. Smith" wrote:
    >> Protocol independence has been the goal of network programming since the
    >> beginning. But for X.25, Sun blithely pushes its NLI interface - which is
    >> basically hard-coded X.25. It shares no abstractions with any other API
    >> and is not compatible with any of the facilities that support all other
    >> protocols as a group. Plus, it is complicated to use (despite that Sun
    >> uses happy tones to market it).

    >
    > My guess is that the X.25 market is small and shrinking rapidly


    No guess required; one of the things I've done at my last 2 employers is
    tearing out X.25 networks and junking them.



    --
    Ignorance more frequently begets confidence than does knowledge: it is those
    who know little, not those who know much, who so positively assert that this
    or that problem will never be solved by science.
    [email me at huge {at} huge (dot) org uk]

+ Reply to Thread