Re: [RE] What's wrong using FireWire as the home backbone?
Comments interleaved ..
On 9/2/04 11:49 AM, "Gross, Kevin" <kevin.gross@CIRRUS.COM> wrote:
>
>> What is missing from FireWire that prevents it from solving the problems
> of wired home networking?
>
> 1/ FireWire does not like to go further than 4.5m
A bit of a misperception ... FW400 (also known as 1394-1995 and 1394a-200)
used 4.5m for the default cable construction. The cable is defined in terms
of performance, not length. Reasonable cables are, however, limited to 10m
or so.
The 1394b-2002 spec (the basis of FW800) defines 100m links using UTP (cat5)
at S100 (98.304Mbit/sec) that are shipping now (in modest quantity ... see
the TI homepage and search for TSB17BA1), and 100m S800 (8x98.304Mbit/sec)
optical links are also being used for professional applications (see
http://www.newnex.com/firenex.html).
> 2/ FireWire does not scale well - fixed amount of bandwidth is shared
> amongst all devices on the network.
It scales fine, and exactly the same way Ethernet does (better, perhaps).
Within a single arbitration domain, the bandwidth is indeed shared ... it
is, however, shared deterministically, and delays are bounded and short. You
can run a 1394 network right up against it's bandwidth limits and nothing
nasty happens (there are no collisions and packets are not dropped).
To get around the shared bandwidth problem (and it *is* a real problem for
some applications), then 1394.1 bridges are required. These are really quite
similar to Ethernet switch/bridges *except* that they preserve the nicely
paced, nicely synchronized world of 1394 isochronous traffic. No products
exist, however, just a number of interesting prototypes.
> 3/ FireWire is not Ethernet
And that *is* the crux of the issue, is it not? So. Let's not argue about
how Ethernet can replace FireWire, or how FireWire can replace Ethernet. In
technical terms (and *purely* technical terms), FireWire has superior
services to Ethernet ... but those services come at a cost, and the cost may
not be worth the trouble.
here are the facts:
1) Ethernet owns the networking world ... yes, yes, there are interesting
and useful exceptions for particular markets.
2) FireWire owns the consumer electronics A/V interconnect space ... yes,
yes, there are places where other things are used (particularly wireless),
but Ethernet is *not* currently in the game.
So. Let's not fight. Let's *merge* the technologies. I *strongly* believe
that Ethernet *can* be the backbone for 1394 local clusters. Sure, the
*technical* performance will not be as good as a pure 1394 network, but I am
not interested in getting into that battle. An enhanced Ethernet will be
*good enough*, and that's just fine.
As far as cost, some preliminary work that I have done has convinced me that
the cost of giving Ethernet adequate isochronous and real-time services
(i.e., comparable with 1394) is *not* significant. There will be no cost
increment at the end-point beyond one set of extra FIFOs (so that real-time
priorities will be respected), and some minor logic enhancements to a
switch.
There *will* be significant and subtle work in the standards arena to make
sure that a good and architecturally consistent specification will be
produced. Fortunately, both Ethernet and 1394 are good, strong, and mature
technologies and there is room to make them work together.
> Items 1 and 2 have been addressed by enhancements to the IEEE-1394
> specification (FireWore over CAT5 and bridging). I'm not seeing quick market
> uptake on these advances. Probably because of item 3. Ethernet is popular.
> Independent of technical capabilities, the more popular an interconnect is
> the more powerful it is. This is a basic property of networks Datacom and
> otherwise.
In general, I agree. The "more popular, the more powerful" statement should
be "the more popular, the more marketing attention it receives, so the more
resources are applied, ..." etc, etc.
(For those that don't know me, be aware that I designed Ethernet into a
distributed PBX that I architected back in the early '80s, helped with the
FDDI-2 design (RIP), and was the tech lead and architect for Apple's
"ChefCat" project ... which became today's 1394 and FireWire. I have about
as much technical, professional, and emotional attachment to FW as anyone
... but I'm also by nature a synthesizer, and I think 1394-over-802.3 is a
natural progression that will bring us one step closer to the "just plug it
in" marketing promise of every network huckster since the '80s.)
(more on my position on these issues can be found in a presentations I made
at the Portland meeting ... see
http://www.ieee802.org/802_tutorials/july04/1394HistoryAndMarket.pdf and
http://www.ieee802.org/802_tutorials/july04/WorkingTogether.pdf
... a quick 1394 technical summary can be found at
http://www.ieee802.org/802_tutorials/july04/NewTechIntroTo1394.pdf and the
1394c UTP/Gigabit summary can be found at
http://grouper.ieee.org/groups/1394/c/1394cIntroKevinBrown.pdf)
--
-----------------------------------------------------------
Michael D. Johas Teener - Mike@Teener.com PGP ID 0x3179D202
23 Acacia Way, Santa Cruz, CA 95062-1313
+1-831-247-9666, fax +1-831-480-5845
------------------- www.teener.com ------------------------