Seeking sustainable IT (not yet desperately, but still ...)

Underlying recent debates I've been involved here and there was a similar question : "Who cares about yet another language specification?". I found myself answering quite at the same time "Yes, please" on one side, and "No, thanks" on the other. And in this latter case, I was striken by a remark from Jim Mason - whose background in standard matters makes me always consider very carefully whatever he brings.
Let's face it. We're building these things for ourselves, and they're proliferating because we have fun doing it.
And any sofware vendor or consultant around could have added : " ... and because we hope to sell more technology build on top of it."
It made me wonder about the relevancy of some of the implicit assumptions which pushed me into Knowledge Engineering quite a while ago, and which I explicited at some point in a nutshell as : "Knowledge is sustainable information". Information is consumable and volatile, will be tomorrow at best redundant, at worst obsolete. By opposition, knowledge is supposed to be sustainable and building up with time. The more knowledge you have already gathered, the more you are likely to transform new information into more knowledge. So I envisioned "Knowledge Technologies" (KT) as another name for "Sustainable IT", along the lines of the European IST program spirit : "From Information Society to Knowledge Society". All of it was supported by my background, which made me consider Maths as the most impressive accumulation of (sustainable) knowledge ever, after natural languages of course.
Right opposite to this cumulative and patient build-up of knowledge we see in Maths and Science, proliferation of technology for the sake of it is clearly anything but sustainable. And actually, the current trend in which languages, software and hardware are all tied up in technological packages leads to this annoying conclusion that languages and specifications are bound to follow the same kind of "product life cycle" logic than their supporting software and harware. If Knowledge Technologies keep up following such a track, they clearly more belong to the technology-for-the-sake-of-it market logic than to sustainable knowledge building.
A very pernicious trend indeed, for many obvious reasons. Beyond the sheer issues of managing semantic interoperability between current, past and future languages and formats, lost of critical knowledge and data embedded in obsolete formats (see e.g. the Pionneer Anomaly), there are the human aspects of it : playing with a language is always more or less formatting your way of thinking. Dealing with too many different languages is difficult and confusing. Concepts are embedded in their representations, so considering language product life cycle means also considering concept life cycle. Not a very pleasant perspective.
I'm not sure what the requirements for sustainable IT would be. But surely one on them would be considering concepts the same way as life forms, with their needed diversity, fragility, and need for care.