Large scale buy-in from the worldwide community of working linguists

If linguists and other language scientists aren't sharing data, the existence of a cyber-infrastructure becomes moot. One of the foremost needs is a culture change within the field that assumes, encourages, and rewards the archiving and sharing of data.

Web-based environment for storage, flexible annotation and analysis of corpus data

I.e. something like NLTK but with a GUI, something like TextGrid, but with a web-based interface instead of Eclipse (Adobe AIR?). A virtual research environment for corpus linguistics that allows data sharing and the use of centralized storage and computational resources.

Language Catalog

A comprehensive language catalog, providing HTTP URLs for languages, dialects, etc would be necessary to integrate linguistic data using linkeddata mechanisms.

A standard for interlinear glosses

A standard way to exchange interlinear glosses would be nice, maybe some sort of XHTML based microformat.

Single user identity ('single logon')

There is ongoing work on a global federation of e-identification providers for research and education such that users can access remote resources with a single login. (See also Kalmar2, EduGain).

Alternatively, a service provider such as linguist list may run an OpenID indentity provider service for linguists. This professional identity could be used across OpenID enabled resources.