Jump to content

Talk:Gigabit Ethernet

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Crossover

[edit]

I have removed some misleading information about crossover cables in 1000Base-T. Somebody should write a better piece about how 1000BaseT has no TX and RX pairs, but gigabit NICs will usually negotiate MDI-X for legacy 100Base-TX connections. 2A00:67A0:A:6:222:15FF:FE88:AAE2 (talk) 16:19, 3 June 2014 (UTC)[reply]

No, not misleading. 1000BASE-T does transmit and receive on all pairs but pairs need to be matched before transmitted data can be properly received. Note that there's no other handshake than Autonegotiate and Auto-MDIX for finding out which pair is what. For 100BASE-TX downward, Auto-MDIX works just the same way as with native FE ports. After all, a GbE port can't really do MDI-X when connected to a legacy switch/hub port, can it? Zac67 (talk) 19:34, 3 June 2014 (UTC)[reply]
As far as I know, all 1000baseT/100baseTX combined ports do automatic MDI-X, and for 1000baseT figure out any possible combination of the pairs. (They do have to be pairs, though.) Since al pairs have both transmit and receive logic, it is easy to do, and they usually do. Gah4 (talk) 05:33, 4 March 2022 (UTC)[reply]
Auto MDI-X 'comes for free' on 1000BASE-T ports (requiring no additional hardware but some logic still), but it's generally optional and not all 10/100/1000 ports support Auto MDI-X. That fact is often hidden by the partner port supporting auto crossover. MDI-to-MDI-X pairing is still required for the Auto Negotiation phase where a physical crossover is always needed. For 100BASE-TX ports, there may be even more not supporting Auto MDI-X from my experience.
My comment above from 2014 isn't quite correct. I've learned since that 1000BASE-T's PCS sublayer matches each lane pair in the link phase. When Auto Negotiation has worked the rest of the pairings shouldn't matter (so a two-pair crossover works with 1000BASE-T as well, without requiring a special four-pair crossover between alike ports). --Zac67 (talk) 06:15, 4 March 2022 (UTC)[reply]
I always made my crossovers of all four pairs, but it seems that the usual is two pairs. As well as I know, the standard doesn't require 10/100/1000 ports to do auto-MDI-X, but it would be a well known deficiency. No-one wants their product to be known as less capable, even if it doesn't matter in actual use. Gah4 (talk) 09:30, 24 December 2022 (UTC)[reply]

Please explain why you reverted my edit to Gigabit Ethernet

[edit]

Thank you for watching over Wikipedia.

You reverted this edit [1] with a comment about it being Too prominent for single sentence entry in a person article. But the edit isn't in a person article, it's in an article about a technical invention (Gigabit Ethernet).

I inserted a single sentence, in a section named "History", because the section never actually said who invented Gigabit Ethernet. The existing sentences were about generic Ethernet, or about the standards committee work that followed the technical invention. A single sentence about the actual history of gigabit ethernet's invention, among seven sentences in the lead paragraph of "History", didn't seem out of place to me. Many of the actors doing that invention already have Wikipedia pages, so are notable, worth mentioning. Two news reports support the facts mentioned. I'm still not sure what you think was too prominent about my edit. Please explain. Gnuish (talk) 06:10, 19 December 2019 (UTC)[reply]

The edit summary was a bit screwed up, sorry. The reasons for revert: Granite Systems doesn't even have its own page, instead it redirects to David Cheriton. Your edit makes this company for too prominent for "a gigabit Ethernet switch start-up" (according to source). If the company or person article has earned that prominence for inventing anything special, being first with anything etc, that isn't established by either source or the article page itself. As it is, there are a dozen companies like that and we don't want to mention them all here. --Zac67 (talk) 08:42, 19 December 2019 (UTC)[reply]
It might be that the 3Mb/s Ethernet that Xerox developed for the Alto was mostly done by one person, but in fact was done by four.[1] But if any one person is named for Gigabit, it would be Rich Seifert who, as they say and that link notes, wrote the book on Gigabit Ethernet. Gah4 (talk) 10:18, 19 December 2019 (UTC)[reply]
@Gnuish: I agree with Zac67. Establishing notability for Granite Systems would be a good place to start; The ref you included with your reverted contribution does not get us there. ~Kvng (talk) 16:02, 3 January 2020 (UTC)[reply]
I do have to say that I have been surprised both ways on what is notable. I will ask Rich and see what he says. Gah4 (talk) 17:52, 3 January 2020 (UTC)[reply]
OK, the official results are in. Rich says:
As far as contributing to the DESIGN (i.e., algorithms, circuitry, encoding, etc.), David did nothing. Most of the work was done under the auspices of the IEEE 802 committee, and (to the best of my knowledge) I never saw David attend a single meeting. I have no idea what he was working on in the background at Stanford (he was a professor there).
and
David DID do work on gigabit switches, and together with Andy Bechtolsheim (one of the founders of Sun Microsystems), founded a company called Granite Systems to capitalize (monetize) their switch and chip effort. Granite was purchased by Cisco (at an outrageous price, as I recall). The purchase was not so much about Granite’s technology, but simply a way to hire an existing team of engineers who already were “on the ground, running.” The acquisition was measured in “X million per employee”.
Also, seems that means that Granite Systems should redirect to Cisco. Gah4 (talk) 20:56, 3 January 2020 (UTC)[reply]
Granite Systems is not mentioned in Cisco Systems so not a good target. We could redirect to List of acquisitions by Cisco Systems. I'm not convinced that's better than the current destination: David Cheriton ~Kvng (talk) 01:04, 9 January 2020 (UTC)[reply]
I asked on talk:Cisco Systems but no reply yet. Gah4 (talk) 04:57, 9 January 2020 (UTC)[reply]

References

  1. ^ "Multipoint data communication system with collision detection". patents.google.com. USPTO. Retrieved 19 December 2019.

Optical interoperability section

[edit]

Essentially, the Optical interoperability section doesn't make much sense:

  • There may be optical interoperability with respective 1000BASE-X Ethernet interfaces on the same link. is misleading – the reference talks about 100BASE-X and is a bit ambiguous even. If interoperability is touched on it needs to be elaborated, but maybe not even here.
  • It is also possible with certain types of optics to have a mismatch in wavelength. without further explanation can be misleading as well. Also, there's nothing special about 1000BASE-X here, optical wavelength spectrum compatibility is a general fiber optic topic and should be discussed there.
  • To have interoperable there is some criteria that have to be meet: is rather obvious but makes very little sense without further elaboration. Again, it's a general Ethernet topic and not specific to fiber or gigabit.
  • 1000BASE-X Ethernet is not backwards compatible with 100BASE-X and is not forwards compatible with 10GBASE-X. - obviously not, as no Ethernet PHY variant is backward or forward compatible with any other.

I don't think there's anything that can be salvaged. Any objections to deleting the section? --Zac67 (talk) 19:33, 28 July 2020 (UTC)[reply]

I think there should be a section, but agree that this is not it. There are some not so obvious combinations of fiber that can be used, other than the one that each was designed for. One can often use multi-mode fiber over short distances with otherwise single-mode interfaces. I didn't look, but I suspect that the fiber article doesn't get into that. Gah4 (talk) 20:57, 28 July 2020 (UTC)[reply]
There is, specifically, a problem with 1000baseLX in multi-mode fiber, slightly described here. Single mode fiber uses a smaller spot, which should not cause a problem. It seems, though, that much multi-mode fiber has a defect in the exact center, which doesn't cause problems with normal use, but does with ones like 100baseLX. The solution is a special patch cable called controlled-launch, which puts the beam slightly off center. One of those strange things that needs to be described somewhere. I am not sure between the different fiber sizes and such, which might also be described here.

line code

[edit]

The copper table has a line code column, but not the fiber table. Can we have one? Gah4 (talk) 05:36, 4 March 2022 (UTC)[reply]

This information is listed in the respective table header already. There is no separate column, because this specification is identical for most entries. Please also note that the table follows a unified formatting throughout several articles which should be preserved. Nightwalker-87 (talk) 01:06, 6 March 2022 (UTC)[reply]

Should data outside of published standards be included in tables?

[edit]

The table in Fiber optics specifies that 1000BASE‑SX can be used at a range of 1km on OM3 grade cable. This is an outlier; all other multi-mode limits are 550m, even on the longer-wavelength 1000BASE‑LX. The OM3 figure is not part of the 802.3z-1998 standard, as OM3 was introduced in 2003. I could not find a source for the 1km figure, and found a conflicting source specifying 860m. If a reliable source can be found, is it ok to include this datum, or should the table not include data outside of the relevant standard at all? Belden.stephen (talk) 21:19, 23 December 2022 (UTC)[reply]

Information outside the standards must be well sourced or should be removed. Perhaps we should mark those claims with {{cn}} before removing them outright. --Zac67 (talk) 21:36, 23 December 2022 (UTC)[reply]
I did further research and found 3 independent sources which verify this claim. It seems to be at least an industry standard, so probably has value remaining in the table. I have added the most reputable of the sources I found. Belden.stephen (talk) 06:13, 24 December 2022 (UTC)[reply]