HUGnetLab has a great advantage that I'm reluctant to abandon. It is built around a net broadcast framework and includes near-real-time graphing and downloadable, persistent .csv data archiving.
It also disadvantages as Ecco has reminded us, but these can be overcome in principle, because it is open-source. Here's a casual design spec (wish list) to open the discussion:
* Flexible, expandable and stable front-end with easy setup and provision for calibration in standard units.
Support for existing standard data interface formats (like GPIB) is essential for integration of 3rd party instruments.
* A/D sampling at 1k samples/sec minimum and configurable per-channel averaging for output data.
* Multiple binary/pulse/counter inputs with time stamp per event and sparse data storage.
* Near-real-time data upload to a public server, with graphic display of data and csv archive.
* Reasonable cost and easy learning curve for end users.
Not asking too much am I?