Monday, 20 April 2009

Network Cable Connectors: Can the design be got right this time?

The shift to 10 Gigabit networking presents an opportunity for the industry standardisation bodies to rectify two glaring inadequacies in the design of the standard data connectors which are based on consumer telephone equipment and as implemented are, by design, clearly unsuitable for professional uses to which they have been put in the case of computer networking since the widespread change from coaxial to twisted pair cable.

The first problem with RJ-45 connectors is that they have a snaplock fitting which breaks off easily. Coming from a background of experience with the XLR and similar pro grade audio connectors which have properly designed snaplocks that don’t break off, and more particularly aren’t prone to getting caught in other cables, clothing or any manner of other things and broken off that way.

The second lesson the IT industry needs to learn from professional audio experience is the matter of the plug gender. For unfathomable reasons, there is only one kind of plug that can be fitted on the RJ 45 patch cable. Every piece of equipment is fitted with female sockets and every patch lead is fitted with male plugs. Thus, you get the situation where any cable can be plugged into any two sockets on the same network switch, creating a loop situation. The pro audio industry, long ago, dealt with their equivalent by reversing the gender of input and output sockets. For example, a signal output on a piece of equipment is typically a female socket, and a signal input is typically a male socket. Inadvertent connections of a cable to the wrong socket are avoided thus. The IT industry should have switches fitted with one gender of connector, NICs fitted with the other type of gender, and cables that have a male plug on one end and a female plug on the other. Thus the common and loathsome cable loop is avoided.

10 Gig will use new types of connectors because the ubiquitous cheap and nasty RJ-45 is inadequate at such speed. I hope the standardisation gurus resist the call for backward compatibility and produce a much better design this time around. Backcompat is actually not very common in the industry (for example, SATA is totally incompatible with IDE, PCI-E is totally incompatible with old PCI) and therefore is not worth bothering with.