Does socket represent an interface between ... ? - TCP-IP

This is a discussion on Does socket represent an interface between ... ? - TCP-IP ; hello 1) a) Does in TCP/IP protocol suite socket represent an interface between Application layer and Transport layer? b) Does in OSI reference model socket represent interface between Application layer and presentation layer? 2) Protocols on application layer can themselves ...

+ Reply to Thread
Page 1 of 2 1 2 LastLast
Results 1 to 20 of 38

Thread: Does socket represent an interface between ... ?

  1. Does socket represent an interface between ... ?

    hello


    1)
    a) Does in TCP/IP protocol suite socket represent an interface between
    Application layer and Transport layer?

    b) Does in OSI reference model socket represent interface between
    Application layer and presentation layer?



    2)
    Protocols on application layer can themselves also be programs ( like
    FTP ), or they can be used by other programs ( SMTP is used by most
    email application ). Now if an email application implements SMTP
    protocol, doesn't that mean that this email application is located
    inside application layer?


    thank you

    cheers


  2. Re: Does socket represent an interface between ... ?

    In article <1181606067.627573.236620@h2g2000hsg.googlegroups.c om>,
    kaja_love160@yahoo.com wrote:

    > hello
    >
    >
    > 1)
    > a) Does in TCP/IP protocol suite socket represent an interface between
    > Application layer and Transport layer?
    >
    > b) Does in OSI reference model socket represent interface between
    > Application layer and presentation layer?
    >
    >
    >
    > 2)
    > Protocols on application layer can themselves also be programs ( like
    > FTP ), or they can be used by other programs ( SMTP is used by most
    > email application ). Now if an email application implements SMTP
    > protocol, doesn't that mean that this email application is located
    > inside application layer?


    Is this homework?

    --
    Barry Margolin, barmar@alum.mit.edu
    Arlington, MA
    *** PLEASE post questions in newsgroups, not directly to me ***
    *** PLEASE don't copy me on replies, I'll read them in the group ***

  3. Re: Does socket represent an interface between ... ?

    kaja_love160@yahoo.com wrote:
    > hello
    >
    >
    > 1)
    > a) Does in TCP/IP protocol suite socket represent an interface between
    > Application layer and Transport layer?
    >
    > b) Does in OSI reference model socket represent interface between
    > Application layer and presentation layer?
    >
    >
    >
    > 2)
    > Protocols on application layer can themselves also be programs ( like
    > FTP ), or they can be used by other programs ( SMTP is used by most
    > email application ). Now if an email application implements SMTP
    > protocol, doesn't that mean that this email application is located
    > inside application layer?
    >
    >
    > thank you
    >
    > cheers
    >


    Here is a question for you.....when is the homework or takehome exam due ?

  4. Re: Does socket represent an interface between ... ?

    On Jun 12, 4:38 am, Justa Lurker wrote:
    > kaja_love...@yahoo.com wrote:
    > > hello

    >
    > > 1)
    > > a) Does in TCP/IP protocol suite socket represent an interface between
    > > Application layer and Transport layer?

    >
    > > b) Does in OSI reference model socket represent interface between
    > > Application layer and presentation layer?

    >
    > > 2)
    > > Protocols on application layer can themselves also be programs ( like
    > > FTP ), or they can be used by other programs ( SMTP is used by most
    > > email application ). Now if an email application implements SMTP
    > > protocol, doesn't that mean that this email application is located
    > > inside application layer?

    >
    > > thank you

    >
    > > cheers

    >
    > Here is a question for you.....when is the homework or takehome exam due ?


    In short, this is not a homework since I'm not a student of any school

    I'm learning about TCP/IP stack and if some questions arise to which I
    don't know an answer to, I hope for someone on an internet to help me
    out and not pest me with silly and non necessary questions ( if this
    was a homework, then it would mean teacher does a piss poor job at
    making up good homeworks ).






  5. Re: Does socket represent an interface between ... ?

    In article <1181606067.627573.236620@h2g2000hsg.googlegroups.c om>,
    wrote:
    >2)
    >Protocols on application layer can themselves also be programs ( like
    >FTP ), or they can be used by other programs ( SMTP is used by most
    >email application ). Now if an email application implements SMTP
    >protocol, doesn't that mean that this email application is located
    >inside application layer?


    No. The protocols and the programs -happened- to be named the same
    thing, but that doesn't mean that they -are- the same thing.

    If you have a dog named Dave, and a cousin named Dave, then does that
    mean that your cousin is a dog?

  6. Re: Does socket represent an interface between ... ?

    wrote:

    > 1)
    > a) Does in TCP/IP protocol suite socket represent an interface between
    > Application layer and Transport layer?


    A socket, IMO, represents the interface between an individual
    application and the Transport Layer.

    > b) Does in OSI reference model socket represent interface between
    > Application layer and presentation layer?


    I'd say no, because the socket is really the interface to the Transport
    Layer. The way I'd say it is that OSI layers 5-7 are all resident in
    what we now call "the application," and that a socket interfaces this
    combined Layer 5-7 blob to the Transport Layer.

    > 2)
    > Protocols on application layer can themselves also be programs ( like
    > FTP ), or they can be used by other programs ( SMTP is used by most
    > email application ). Now if an email application implements SMTP
    > protocol, doesn't that mean that this email application is located
    > inside application layer?


    I'd say that the interface to the e-mail application is at the
    Application Layer, and that you can have multiple such applications open
    at the same time, through different sockets.

    Bert


  7. Re: Does socket represent an interface between ... ?

    thank you

    cheers


  8. Re: Does socket represent an interface between ... ?

    > wrote:

    >> b) Does in OSI reference model socket represent interface between
    >> Application layer and presentation layer?

    >
    > I'd say no, because the socket is really the interface to the
    > Transport Layer. The way I'd say it is that OSI layers 5-7 are all
    > resident in what we now call "the application," and that a socket
    > interfaces this combined Layer 5-7 blob to the Transport Layer.


    Let me be more exact.

    If you want to stick literally to the OSI model, then I'd say that a
    socket represents the interface between the Transport Layer, e.g. TCP,
    and the Session Layer, which launches and controls each application that
    is currently running over that IP link. I suppose just because Layers
    5-7 are shrink-wrapped in the same package does not mean that
    conceptually they don't exist.

    The Presentation Layer is above this. The socket doesn't care about the
    details such as scaling of variables, binary or ASCII, internal headers,
    delimiters, etc.

    Bert


  9. Re: Does socket represent an interface between ... ?

    kaja_love160@yahoo.com writes:

    > On Jun 12, 4:38 am, Justa Lurker wrote:
    >> kaja_love...@yahoo.com wrote:

    >
    > In short, this is not a homework since I'm not a student of any school


    Well, OK. Your questions seem just far enough off the wall that they
    struck many of us as seeming more like a professor testing someone's
    knowledge some someone learning for themselves.

    >> > a) Does in TCP/IP protocol suite socket represent an interface between
    >> > Application layer and Transport layer?


    There are two kinds of "socket" here, and I'm not sure which you are
    referring to. "In TCP/IP protocol suite" -- i.e., in the TCP/IP
    protocol specifications -- there is a concept of socket that
    represents the bidirectional interface between two peer applications
    on two different nodes. The application on one node puts data into its
    end of the socket, and its peer pulls that data out of its end. In
    that context, the interface between the application and the transport
    protocol implementation is something else; in the TCP/IP spec, that
    something else is a handful of functions.

    In UNIX (and, now, most everywhere), there is an API mechanism that is
    also called a "socket". (I've always assumed the UNIX name grew out of
    the TCP/IP name, but I don't really know for sure.) This "socket"
    provides the application with a handle *to* the protocol suite's
    "socket", and also provides a handle for related activities such as
    setting sockets up and relating them to the local I/O system. I think
    it would be reasonable to call the UNIX "socket" an interface between
    the application and the transport protocol.

    By the way, the OSI model doesn't really work that well when cut so
    finely in the modern world. For example, TCP/IP, as an OSI "layer", is
    somewhere in the middle of session layer, so half of session,
    presentation, and application levels are all above it.

    >> > b) Does in OSI reference model socket represent interface between
    >> > Application layer and presentation layer?


    It's been so long since OSI was of any relevance, so my memory is a
    little fuzzy, but as I recall TP-4 used the term "socket" or something
    similar for the same peer-to-peer communication path as the TCP
    socket. And TP-4 has been implemented under UNIX sockets, so its
    relation to them is the same. (Back in the day, though, TP-4 was
    normally discussed in an I/O system called "Streams", and so a
    "stream" was the more common API handle discussed....until everyone
    realized OSI was guff and abandoned it and all the other
    new-but-not-really-new artifacts that had grown up around it such as
    Streams.)

    >> > 2)
    >> > Protocols on application layer can themselves also be programs ( like
    >> > FTP ), or they can be used by other programs ( SMTP is used by most
    >> > email application ). Now if an email application implements SMTP
    >> > protocol, doesn't that mean that this email application is located
    >> > inside application layer?


    Do not confuse applications with Application Layer. This is one thing
    OSI actually got right. Application Layer describes application
    *protocols*, such as the OSI equivalents of SMTP and FTP. *On top of*
    the Application Layer ran *an* application that actually provided a
    functional service to its customer. (Some people called this "an
    eighth layer": the user interface.) OSI was actually very big on
    imagining applications which used multiple Application Layer Protocols
    to implement Really Cool applications beyond what the Application
    Protocol designers had imagined: A Dream of Emergence.

    As someone already pointed out, in the TCP/IP this can be a little
    confusing because what you would call "application protocols" in
    TCP/IP are normally very directly functional, so the first
    implementation often is the Application Protocol with a little, tiny
    "application", really no more than a way for human input to be
    converted into protocol bits, wrapped together into a single program
    that has the name of the protocol. To get a better understanding of
    the actual applications as contrasted with Application Protocols,
    think of applications like mail systems or name resolutes systems that
    use many different Application Protocols (often by invoking those
    simpler programs in subprocesses) from many different protocol suites
    to accomplish their missions.

    -don

  10. Re: Does socket represent an interface between ... ?

    "don provan" wrote:

    >>> > b) Does in OSI reference model socket represent interface between
    >>> > Application layer and presentation layer?

    >
    > It's been so long since OSI was of any relevance, so my memory is a
    > little fuzzy, but as I recall TP-4 used the term "socket" or something
    > similar for the same peer-to-peer communication path as the TCP
    > socket.


    Just as an aside, I think the OP (original poster) was asking about
    typical UNIX sockets, and I think that the 7-layer ISO/OSI model is just
    as valid for IP as it was for ISO protocols. As a conceptual tool, it
    works well in both cases.

    We have been over this topic in this usenet group before. I think that
    while the names might be a little different, and while OSI is a 7-layer
    model vs a 4-layer model used in early RFCs, there really is no big
    disconnect there. ISO protocols might have been abandoned, but the
    7-layer OSI model is as good as any conceptual model for digital comms,
    IMO.

    That aside, I agree with your points.

    Bert


  11. Re: Does socket represent an interface between ... ?


    "Albert Manfredi" writes:
    > Just as an aside, I think the OP (original poster) was asking about
    > typical UNIX sockets, and I think that the 7-layer ISO/OSI model is
    > just as valid for IP as it was for ISO protocols. As a conceptual
    > tool, it works well in both cases.
    >
    > We have been over this topic in this usenet group before. I think that
    > while the names might be a little different, and while OSI is a
    > 7-layer model vs a 4-layer model used in early RFCs, there really is
    > no big disconnect there. ISO protocols might have been abandoned, but
    > the 7-layer OSI model is as good as any conceptual model for digital
    > comms, IMO.


    OSI model lacks any internetworking layer ... that was one of the
    problems with OSI ... aggrevating the situation was that ISO had
    procedures that precluded working on any (networking) standards that
    failed to conform to OSI model ... which precluded anything that
    supported internetworking ... and for that matter for anything that
    supported LANs/MACs; aka LANs/MACs have an interface/api that sits
    somewhere in the middle of layer3/networking (above layer2/layer3
    interface but below the layer3/layer4 interface) ... and the
    internetworking layer (ability to internetwork multiple networks)
    doesn't exist at all in OSI.

    part of arpanet and OSI evolving thru the 60s and 70s was support of a
    homogeneous network (i.e. network of networking nodes). so some number
    of the IETF RFCs reflect the homogeneous arpanet/OSI type of
    approach. However, within the IETF community in the 70s ... it was
    realized that networking wasn't going to be sufficient and work started
    on internetworking. about the time that ISO finally passed the initial
    OSI standards in the early 80s ... IETF was converting from homogeneous
    kind of arpanet networking to internetworking (the big 1jan83 conversion
    to tcp/ip)

    then thruout the 80s, OSI continued to hang on ... somewhat as the
    difference between networking and internetworking started to slowly
    permeate the conscience of wider community (even tho still in the 1990
    time-frame there were still various gov. mandates to eliminate the
    internet and have it replaced with ISO/OSI).

    possibly, part of the issue of looking at TCP/IP compared to OSI at a
    purely protocol level ... is that the complexities of internetworking
    are as much at the operational and business levels ... as the
    technical/protocol level (altough the finer nuances of internetworking
    technical/protocol have to be in place to enable the operational and
    business caracteristics).

    misc. past posts mentioning OSI and/or attempt at high-speed protocol
    standardization effort in ISO ... which was precluded based on violating
    OSI model (aka support for both LAN/MAC as well as tcp/ip
    internetworking).
    http://www.garlic.com/~lynn/subnetwork.html#xtphsp

    a few recent posts discussing 1) tcp/ip being the technology basis for
    modern internet(working), 2) NSFNET backbone being the initial
    operational basis for modern internet(working) and 3) early CIX
    (commercial interchange) being the business basis for modern
    internet(working)
    http://www.garlic.com/~lynn/2007d.html#43 Is computer history taugh now?
    http://www.garlic.com/~lynn/2007h.html#38 sizeof() was: The Perfect Computer - 36 bits?
    http://www.garlic.com/~lynn/2007i.html#69 How the Internet took over
    http://www.garlic.com/~lynn/2007j.html#70 Using rexx to send an email
    http://www.garlic.com/~lynn/2007l.html#37 Friday musings on the future of 3270 applications
    http://www.garlic.com/~lynn/2007l.html#67 nouns and adjectives
    http://www.garlic.com/~lynn/2007l.html#68 nouns and adjectives
    http://www.garlic.com/~lynn/2007l.html#69 nouns and adjectives
    http://www.garlic.com/~lynn/2007m.html#1 nouns and adjectives
    http://www.garlic.com/~lynn/2007m.html#7 nouns and adjectives

  12. Re: Does socket represent an interface between ... ?

    "Anne & Lynn Wheeler" wrote:

    > OSI model lacks any internetworking layer ...


    You actually made that same comment last time around, IIRC, but it's not
    so. The OSI Network Layer, Layer 3, *is* what early RFCs such as RFCs
    791 and 793 Figure 1 in both cases, call Internet Protocol layer. Also
    RFC 1112, Section 5. In each case, this Network Protocol ties together
    different local area networks, of any type.

    Other early RFCs called this layer the "internetworking layer," and
    showed how it could allow communications between different "networks,"
    which in that context meant LANs, or Layer 2 nets, as we might call them
    today.

    See:

    http://www.ussg.iu.edu/usail/network...rk_layers.html

    for a concise definition of each ISO/OSI layer. That same site makes
    this comment:

    "Although the OSI model is widely used and often cited as the standard,
    TCP/IP protocol has been used by most Unix workstation vendors. TCP/IP
    is designed around a simple four-layer scheme. It does omit some
    features found under the OSI model. Also it combines the features of
    some adjacent OSI layers and splits other layers apart. The four network
    layers defined by TCP/IP model are as follows."

    By the way, the IEEE uses the same 7-layer model, today. It is used in
    their 802 series protocol standards.

    > that was one of the
    > problems with OSI ... aggrevating the situation was that ISO had
    > procedures that precluded working on any (networking) standards that
    > failed to conform to OSI model ... which precluded anything that
    > supported internetworking


    Again, I completely disagree, since the OSI protocols were running
    between FDDI and Ethernet link layer nets quite happily, just as IP was
    and is.

    But more importantly, even if this were the case, it still would not
    make the OSI 7-layer conceptual model any less relevant. The original
    4-layer Internet Protocol depiction is just a simplifed version of that
    OSI model, for all practical purposes. Both ISO and IP were intended to
    allow routing of datagrams between different link layer nets, globally.

    Bert


  13. Re: Does socket represent an interface between ... ?

    >> OSI model lacks any internetworking layer ...

    By the way, the ISO internetworking protocol was called CLNP
    (Connectionless Network Protocol), its interior gateway routing protocol
    was called IS-IS (intermediate system-intermediate system), its exterior
    gateway routing protocol was IDRP (inter-domain routing protocol), and
    Transport Layer protocols were TP4 and CLTP, corresponding to TCP and
    UDP respectively. I think there were other connection-oriented TPs as
    well, but TP4 seemed the one everyone seemed interested in at the time.

    Believe it or not, IS-IS still exists, RFC 1195.

    So I don't see that there were any functional differences in the most
    fundamental sense, between ISO and IP. They had the big NSAP addresses,
    which would have been wide enough to take away a lot of the steam IPv6
    has today.

    Bert


  14. Re: Does socket represent an interface between ... ?

    "Albert Manfredi" writes:
    > You actually made that same comment last time around, IIRC, but it's
    > not so. The OSI Network Layer, Layer 3, *is* what early RFCs such as
    > RFCs 791 and 793 Figure 1 in both cases, call Internet Protocol
    > layer. Also RFC 1112, Section 5. In each case, this Network Protocol
    > ties together different local area networks, of any type.


    Old email from long ago and far away. x3s3.3 was ANSI chartered ISO
    standards group responsible for "OSI" level 3&4 standards work.

    From: wheeler
    Date: 27 Mar 89 21:41:07 GMT

    Quicky note on ansi x3s3.3 and hsp meetings last week. More
    information coming as time allows.

    A "high speed networking & transport protocol" proposal was submitted
    at the x3s3.3 meeting. After various discussions it was decided to
    submit a "study proposal for high speed protocols" to the x3 committee
    .... the work product of which will be some number of protocol
    proposals.

    Problems with the original protocol proposal were numerous. Many
    people objected to it violating the OSI reference model (and in fact
    it is not possible to submit a protocol proposal to X3 that violates
    the reference model ... although it is possible to approve an ANSI
    standard that does violate the reference model ... but that takes some
    fine work ... case in point are the LAN protocols ... especially with
    LAN/MAC coming up thru level 1 and 2 well into level 3).

    .... snip ...

    i.e. it has been possible for ANSI (and/or IEEE) to pass a standard
    (like the 802 stuff) that violate OSI (just that they couldn't do work
    on such standardization violating OSI within ISO chartered group...
    but it wasn't possible to have a standard work item accepted for
    standards work; interesting distinction). x3s3.3 had to object to the
    hsp work item on grounds that it violated OSI ... 1) supporting
    internetwork protocol, 2) going directly from transport to LAN/MAC,
    bypassing level 3/4 interface, and 3) supporting LAN/MAC interface.

    and for past light hearted post (by somebody else) in this n.g. also
    long ago and far away ... but should also be in one of the online
    archives

    Newsgroups: comp.protocols.tcp-ip,comp.protocols.iso
    Date: 1 Apr 88 00:00:01 GMT
    Organization: Soviet Sanctuary for Victims of American Persecution
    Posted: Fri Apr 1 00:00:01 1988

    WASHINGTON -- In a simultaneous announcement that took the
    computer industry by surprise, OSI leaders today said that they were
    abandoning their effort to promote the OSI Protocol Suite in favor of
    the existing US Department of Defense (DoD) ARPANET Protocol Suite.

    The official reason cited for the decison was a new report from
    the Office of Technology Assessment stating that the manpower required
    to fully implement and test even the few OSI protocols that are now
    defined would consume the entire output of American university
    computer science programs for the rest of the century, and that
    printing and distributing the necessary protocol specifications would
    consume the entire American and Canadian paper supplies for the next
    five years.

    However, one high-placed source speaking on condition of anonymity
    said, ``The whole OSI thing was a practical joke one of the guys
    cooked up a few years ago. Nobody ever expected anybody to take it
    seriously. I mean, who would believe an organization supposedly
    dedicated to tearing down barriers to free and open communications
    between computers when it's run by a former director of the National
    Security Agency? I guess computer people are a lot more gullible than
    we thought. We kept dropping hints, making the whole thing more and
    more ridiculous. We hoped that people would eventually catch on, but
    it didn't work. Finally, our consciences got to us.''

    In related news, officials at the Mitre Corporation in Bedford,
    Massachussetts reported that one of their employees, as yet publicly
    unidentified, froze ``as solid as stone'' when he heard the
    announcement. Medical experts have as yet been unable to communicate
    with the victim or get him to relax his facial muscles, which are
    reportedly locked into what was described as an ``enormous grin''.

    .... snip ...

  15. Re: Does socket represent an interface between ... ?

    On Jun 13, 7:57 pm, Anne & Lynn Wheeler wrote:

    > Old email from long ago and far away. x3s3.3 was ANSI chartered ISO
    > standards group responsible for "OSI" level 3&4 standards work.
    > and for past light hearted post (by somebody else) in this n.g. also
    > long ago and far away ... but should also be in one of the online
    > archives


    > Newsgroups: comp.protocols.tcp-ip,comp.protocols.iso
    > Date: 1 Apr 88 00:00:01 GMT
    > Organization: Soviet Sanctuary for Victims of American Persecution
    > Posted: Fri Apr 1 00:00:01 1988


    That was an April Fools joke, but I remember when it actually
    happened. IIRC, it was 1996.

    In any event, no one is arguing that ISO did things efficiently. They
    spent more time writing perfect standards and less time actually
    playing with them to make them work. And the IETF at the time was the
    other way around.

    But this has nothing to do with the fact that ISO and IP were tackling
    the same problem and in much the same way. Both had routing protocols,
    both had internetworking layers, both tied together disparate LAN
    protocols, both used connectionless datagrams to carry both
    connectionless and connection-oriented sessions.

    So in terms of the layered protocol model, the ISO one was simply
    "more complete" than the IP one, but both the 4-layer IP model and the
    7-layer OSI model addressed exactly the same problems.

    The layered model that survived to this day is the 7-layer model. It's
    that simple. Those who slavishly had to follow the 7-layer model to
    the letter, when writing software, were overtaken by events.

    Bert


  16. OSI abandoned!

    Albert Manfredi writes:
    > That was an April Fools joke, but I remember when it actually
    > happened. IIRC, it was 1996.


    re:
    http://www.garlic.com/~lynn/2007m.html#24 Does socket represent an interface between ... ?
    http://www.garlic.com/~lynn/2007m.html#25 Does socket represent an interface between ... ?

    i pulled the copy out of my email archive ... ... as i mentioned in the
    previous post, should be able to find it in one of the online usenet
    archives. here is version ... posted Mar 31 1988, 7:00 pm
    http://groups.google.com/group/comp....3b59f7f112234b

    so are you referring to a warmed version of the above nearly a decade
    later? ... or when some specific party/organization ... taking nearly
    another decade to actually accept it.

    for other topic drift ... various old email mentioning NSFNET backbone
    activity thru the 80s
    http://www.garlic.com/~lynn/lhwemail.html#nsfnet

  17. Re: Does socket represent an interface between ... ?

    "Albert Manfredi" writes:

    > Just as an aside, I think the OP (original poster) was asking about
    > typical UNIX sockets...


    Probably true, which was all the more reason to stress that UNIX
    sockets are not "in the TCP/IP protocol" but are only an
    implementation detail in one class of systems.

    > and I think that the 7-layer ISO/OSI model is
    > just as valid for IP as it was for ISO protocols. As a conceptual
    > tool, it works well in both cases.


    Well, I really disagree. First of all, let's split the 7-layer model
    into its two parts. The lower 4 layers were already well established,
    so on the surface all the OSI model did there was lay out existing
    layers and give them names. This is about all the whole thing is
    really good for: all networking classes teach the 7-layer model, so if
    I ask someone about the datalink or networking layer, they have an
    idea what I'm talking about. But even here, it's only useful at a
    superficial level: if you look at the *details* of the 7-layer model,
    it starts to teach you that certain things in the TCP/IP protocol
    suite cannot exist. ARP, the TCP pseudo header, and TCP's graceful
    close are three examples. (I'll leave it as an exercise to the reader
    to explain how.)

    But let's not forget that it's a 7-layer model, not a four layer
    model. Do you know what happened to the top three layers? The punch
    line is that the upper layers of the OSI protocol suite *don't fit*
    into the session, presentation, and application layers! Specifically,
    they discovered that these functions don't really work as "layers" at
    all, so they basically turned them on their sides. If anyone had taken
    what they actually designed and rewrote the "7-layer model" to fit it,
    it would be "the 5-layer and 2-column model" or possibly the "4-layer
    and 3-column model".

    In short, a lot of it really wasn't right, and even some that was
    logically reasonable is now, given what actual "network layer" and
    "transport layer" and "application layer" protocols have been
    implemented, inaccurate in practice. It has played a role in providing
    a universal language, particularly in the lower layers that people are
    actually taught, but in its absence, I think some other, better model
    would have grown into that role in the same way Knuth became the
    standard for algorithms.

    > We have been over this topic in this usenet group before. I think that
    > while the names might be a little different, and while OSI is a
    > 7-layer model vs a 4-layer model used in early RFCs, there really is
    > no big disconnect there. ISO protocols might have been abandoned, but
    > the 7-layer OSI model is as good as any conceptual model for digital
    > comms, IMO.


    I know what you're saying, and I respect your bravery in defending the
    poor thing. It certainly has been a boon to networking instructors the
    world over who pin their lesson plans to. But I think that the degree
    to which the OSI model has been useful is exclusively the degree to
    which it documented the existing practice of the time. But it went
    beyond that, and, for the most part, the rest of it is of little or
    no value.

    > That aside, I agree with your points.
    > Bert


    Well, thanks, Bert! And our disagreement about the 7-layer model is of
    minor importance, since we're stuck with it, anyway, and it's not
    really doing anyone any harm.

    -don

  18. Re: Does socket represent an interface between ... ?

    In article ,
    don provan wrote:

    > "Albert Manfredi" writes:
    >
    > > Just as an aside, I think the OP (original poster) was asking about
    > > typical UNIX sockets...

    >
    > Probably true, which was all the more reason to stress that UNIX
    > sockets are not "in the TCP/IP protocol" but are only an
    > implementation detail in one class of systems.


    Except that "one class of systems" has effectively become "most classes
    of systems". Windows adopted sockets as its API. Sockets are in POSIX.
    I suspect a number of embedded system OSes have adopted it as well.
    I'll bet its available for VMS. It's pretty much become the de facto
    standard networking API.

    --
    Barry Margolin, barmar@alum.mit.edu
    Arlington, MA
    *** PLEASE post questions in newsgroups, not directly to me ***
    *** PLEASE don't copy me on replies, I'll read them in the group ***

  19. Re: Does socket represent an interface between ... ?

    In article ,
    Barry Margolin wrote:

    >> Probably true, which was all the more reason to stress that UNIX
    >> sockets are not "in the TCP/IP protocol" but are only an
    >> implementation detail in one class of systems.

    >
    >Except that "one class of systems" has effectively become "most classes
    >of systems". Windows adopted sockets as its API. Sockets are in POSIX.
    >I suspect a number of embedded system OSes have adopted it as well.
    >I'll bet its available for VMS. It's pretty much become the de facto
    >standard networking API.


    That's true in the bad old sense of "The great thing about standards
    is that there are so many to choose from."

    Judging from my continuing experience with code that I claim works on
    a variety of versions BSD/OS, FreeBSD, NetBSD, OpenBSD, several flavors
    of Linix, AIX, Solaris, HP-UX, IRIX, OSF1, OpenUNIX, Mac OS/X, WIN32,
    and others, Windows' sockets are only somewhat related to UNIX sockets.
    Writing code that works on both WIN32 and UNIX-like systems is a royal
    pain. Common UNIX idioms such as select() and names of errno's do not
    port to Windows unless you restrict yourself to what might be called
    "pidgin sockets." Even among UNIX-like systems, there are plenty of
    important but incompatible socket facilities including those for getting
    IP (not to mention Ethernet or other link layer) addresses of interfaces.


    Vernon Schryver vjs@rhyolite.com

  20. Re: Does socket represent an interface between ... ?

    Barry Margolin writes:

    > Except that "one class of systems" has effectively become "most classes
    > of systems". Windows adopted sockets as its API. Sockets are in POSIX.
    > I suspect a number of embedded system OSes have adopted it as well.
    > I'll bet its available for VMS. It's pretty much become the de facto
    > standard networking API.


    No argument here. But it is, none-the-less, an implementation detail,
    not part of the protocol.
    -don

+ Reply to Thread
Page 1 of 2 1 2 LastLast