I had always understood (and the Internet has it spewed in multiple places) that if I connect T568A on one side, and T568B on the other I am creating a crossover connection.
[...]
When I look at pinout diagrams showing T568A and T568B, the odd number pins are all color/white with the even number pins solid color. When I look at crossover cables, this is not correct [..] Assuming (1) above is correct, then I don't see how this can be correct.
Both kinds of crossover are correct, but for different Ethernet standards. Your understanding is 'correct' for 10/100 Mbps links, while the cables you're looking at are 'correct' for 1 Gbps links.
Note that T568A⇆T568B only swaps two pairs. That is "crossover" in 10BASE-T and 100BASE-TX only (i.e. 10/100 Mbps Ethernet), as those standards only use two pairs at all, so the crossover simply swaps the 'transmit' and 'receive' pairs leaving the other (unused) pairs untouched. The polarity is preserved, so if pins 1 and 3 (Tx+ and Rx+) were both 'white' before, they both remain 'white' after, and of course the two unused pairs also remain the same – so the stripe/color pattern is preserved in both ends.
(The table is from IEEE 802.3-2012, which is the actual Ethernet specification. It's available for free download if you register an account.)
Gigabit Ethernet, however, is wired differently – instead of separate 'transmit' and 'receive', it now has four 'bidirectional' pairs. The standard for 1000BASE-T defines crossover wiring as swapping all four pairs, which is no longer the same as simply T568A-to-B – in addition to swapping 1+/2- with 3+/6- as usual, you also swap 4+/5- with 7+/8-:
(Table from section 40.8.2 "Crossover function" of IEEE 802.3-2012.)
Polarity is still preserved in Gigabit crossover, but if you look closely you'll see that the wire colors no longer match the polarity – only three pairs follow the 'white=positive, colored=negative' pattern, but pins 4/5 are the opposite. So with a cable that implements full crossover for 1000BASE-T, swapping pairs 4/5 with 7/8, the stripe/color pattern indeed goes out of order.
Gigabit crossover cables are also 10/100 Mbps crossover cables, but not the other way around. (But in practice, I think Gigabit ports compensate for the use of halfway crossover cables?)
I further understand that if plugging into a gigabit switch this will still work because it does crossover correction. Is this understanding correct?
Yes, practically all Gigabit ports and most 100 Mbps ports – not just switches but also end-devices – implement autonegotiation to determine whether they should apply internal crossover or not. This works both ways, so a normal cable will work where crossover is needed, and vice versa.
As mentioned earlier, Gigabit crossover is different from 100Mbps crossover, however, 1000BASE-T specification does talk about automatic detection of any combination of swapped and/or inverted pairs; I am not sure how far this is actually implemented, but in practice I think most Gigabit ports do compensate for "halfway crossover" cables.
I hypothesize maybe there is some differential signalling going on to allow it to work, and the pair orders don't make a difference
Gigabit Ethernet uses balanced differential signalling. On top of that, all pairs are bidirectional. The specification mentions that the signalling scheme is designed to automatically correct any kind of pair swap and/or polarity swap:
I assume this is what allows halfway-crossover cables to still work with Gigabit ports.
(The older 10/100 Mbps standards, if I understand the terminology correctly, use 'balanced' but not 'differential' signals.)
Can anyone confirm that the LED colors are standard across most switches, and that an Orange only light (ie green not on) means there is a problem at the hardware (probably cabling) level?
They aren't.
Commonly though just one LED is used to indicate a working link – a fairly common pattern is that green means "link up at maximum speed" while orange means "link up at lower-than-max", e.g. on a Gigabit switch, it might indicate that only a 100 Mbps link was established. But that is not strictly a standard.
I don't think I've seen switches explicitly report "bad cable" via LEDs, but if both ends are supposedly Gigabit-capable devices, then a "100 Mbps" indication would definitely indicate a problem. Of course, if the cable in fact goes to a 10/100 Mbps port, then this is fine.
If I remember correctly, TL-SG108PE is partially a managed switch – you can access it via IP, and its web interface will show you the actual link speed, and there is a "cable test" feature as well. So just look at the "Port Status" table to see what's up.
In general, I would recommend not making any crossover cables at all, unless you have a very good reason to make one (e.g. if you're working with very old 10 Mbps hardware).
Nearly all Gigabit ports and most 100 Mbps ports will accept straight-through cables in any situation, so just pick one of T568A or T568B and use it everywhere.
Also, I would suggest finding a switch that has a built-in "cable test" feature – many managed switches include it (sometimes named 'VeriPHY'). Usually this feature uses TDR, which has some advantages over the "blinky LED" cable testers, as it can tell you which end of the cable is incorrectly terminated – and it doesn't need any special device to be connected on the distant end. It looks like even TL-SG018PE might have this function.
