Curious Marc - HP LED Display Programmer

Made in the USA. Broadcom still make these 8 character displays and you can get them here for £47.35 each. They draw 250mA per display and have 16 user-definable characters stored in flash memory. This is the programmer for them, so the military can hack their own fonts, I guess.. The datasheet has the built-in character definitions, artistically laid out:

See Juan Browne on the Air India 787 Preliminary Accident Report and Hitachi HD44780U LCD Display Fonts.

Here's their HP 9825 Repairathon.

Subscribe to Curious Marc.

I've been thinking a lot about abstraction in documentation recently. It is really frustrating to me that companies like Intel and ARM insist on producing typeset PDF files of their manuals. These are all in some metalanguage that has to be translated. The metalanguage is English because these were originally American and British companies,  respectively, but the actual data the manuals contain is language-neutral. The task of the technical authors is focussed on producing information in the form of English text and diagrams in Portable Document Format, and making sure that this aligns with the actual physical devices produced in the manufacturing process, i.e. the data.

Take web sites for example. These are systems that are self-documenting, because the world wide web is meant to be a way to represent data in different clients that connect to the different servers. So a web site is a program that generates information in the form of an HTML representation of the data on the server. In the case of the web sites of Intel and ARM, for example, that information includes an HTML representation of the same data that the PDF files describe. The web server generates the HTML and sends it to the clients, which are web browsers, and the web browsers display the data and allow the user to interact with it in some way. These interaction events go back and forth between the servers and the browsers. An enormous amount of people's time is spent in this system of action/reaction events, and it is crucial that the information presented to the user corresponds to the data present on the server.

It seems obvious to me then that what we, as developers of information and communications systems,  should be focussed on is the process of translation of representations of data, not the data, which is continually changing as the world develops, nor the information, which needs to be presented in different ways to different people at different times and in different contexts. For example, someone writing an assembler/disassembler for ARM and Intel CPUs needs to take abstract descriptions of the instruction encoding and map those to assembler syntax. These mappings clearly must be the same as the ones described in the documents available from the web site. So generating the code for an assembler/disassembler is just another representation of the same data. This goes on right through the whole process, because ultimately the chip tapeout needs to be mapped to that same data.

See Peter Wang on How to Democratise AI.

And blogger keeps suggesting I add Google search to what I write. A box keeps coming up telling me that it has identified 6, 8, 10, ... different phrases which could be used as search terms, and offers to insert Google search links to the text I am writing. I never do, but the box still keeps popping up like a child desperate for attention. I did an experiment: see Testing Blogger's Automatic Google Search Term Insertion Option.