Talk:Manufacturing data

From OLPC
Revision as of 01:57, 14 June 2007 by Xavi (talk | contribs) (Reverted edits by 61.238.5.79 (Talk); changed back to last version by 209.6.228.166)
Jump to navigation Jump to search
  • Mitch Bradley approves with the qualification: All items marked TBD must be resolved.
  • Ivan approves, same qualification as above.
  • Ray <approval needed>
  • Ron <approval needed>
  • Richard <approves>
  • Jim <approval needed> comments:

Add to goals: flexibility; there will be many keyboard variants, possibly a number of other variants of the base machine, with time.

I think either only text should be allowed, or if non-text values are essential and allowed, a type field should be added to the tags.

Should text be UTF-8, rather than ascii?

It might be better to use UTF-16 rather than either UTF-8 or acsii. This would mean that every character from the base plane of Unicode could be expressed using exactly two bytes and every character from any other plane of Unicode could be expressed using exactly four bytes. Maybe the tags should be 16 bytes each? Eight bytes for four UTF-16 Unicode characters, six bytes for three Unicode digits (U+0030 to U+0039) in UTF-16 format, to express a data length of 0 .. 999 words each of two bytes, and two bytes for a checksum of the tag. The data section would also have an additional two bytes as a checksum of the data section. This would take more bytes for tags and maybe more bytes for the data depending upon the nature of the data. However, this is a "design it now, apply it for a long time" situation so maybe it is worth doing.

Jim: no one uses UTF-16 anymore. We *don't* want to carry extra conversion tables and code for another unicode format, when UTF-8 is already there. If we do Unicode at all, best it be the same everywhere. The tags can easily be restricted to be ASCII and stay one byte (and ascii is part of UTF-8; not true for UTF-16).


The wireless mac number: I presume this is the only location for the MAC address: there is no mac address in the wireless logic itself? I don't believe in dupicating information that may become inconsistent. What is the format of this string? (presuming it is a string).

I'm not sure language is appropriate in the area at all. What is this going to be used for?

The software load has a default locale already, and multiple languages may be supported by a single software load. We've been working hard to not have to localize the BIOS user-interface by using iconic representations, both to avoid the manufacturing headaches and because the machines are often being used by kids/parents not yet literate.

Keyboard types, however, are very important to know (and while are usually related to language, they aren't necessarily a one-to-one mapping). There are actually 2 pieces to this: the physical layout of the keys (physical size and position of keyboards), of which there will be relatively few variations (but there will be some variation), and how the keys are labeled (a silkscreening like process). There will be many, many of those. But you want to have both pieces of information.

Include calibration data for "everything analog" too?

if they are tested anyway it might make sense to include:

  • Calibration data for battery voltage and current measurement
  • Sensitivity/output power/frequency deviation of WLAN
  • Write speed to NAND (sorry I maybe completely off track but some internally timed flash devices exhibit an increased programming time toward end-of-life. No idea whether this would be available on the NAND used)
  • Display brightness (RGB)
  • Reading of MIC input in Analog Input Mode
  • Touchpad calibration
  • Temperature at which the tests were performed


Consistent date format within tags, YYYYMMDD or YYYY-MM-DD prefered?

The dates included within the tags seem to use two different date formats:

T#: 20061113-B001
SD: 14/11/2006

Please go for YYYYMMDD or YYYY-MM-DD (ISO8601). Some reasoning given there: http://www.uic.edu/depts/accc/software/isodates/index.html --Frief 08:18, 20 November 2006 (EST)

Table from Trac bug 435

Here is the table from David Woodhouse:

System SKU B1-1 B1-2 B1-3 B1-4 B1-5 B1-6 B1-7 B1-8 (special hinge) B1-9
1st level PN (5 in 1) 1CL1XZU0KD0 1CL1XZU0KDB 1CL1XZT0KD0 1CL1XZT0KD1 1CL1XZP0KD0 1CL1XZP0KD1 1CL1XZ-0KD0 1CL1XZ-0KD1 1CL1XZQ0KD0 1CL1XZQ0KD1 1CL1XZABKD0 1CL1XZABKD1 1CL1XZQ0KD0 1CL1XZQ0KD1 1CL1XZU0KD6 1CL1XZU0KD1 1CL1XZU0KD6 1CL1XZU0KD1
Packing 5 in 1 Single 5 in 1 Single 5 in 1 Single 5 in 1 Single 5 in 1 Single 5 in 1 Single 5 in 1 Single 5 in 1 Single 5 in 1 Single
DRAM 256 256 256 256 256 256 256 256 256
3rd level PN 31CLMB0002 31CLMB0002 31CLMB0002 31CLMB0002 31CLMB0002 31CLMB0002 31CLMB0002 31CLMB0002 31CLMB0002
K/B language English (USA) Portuguese Spanish Thai Arabic Hausa;Igbo Arabic English (USA) English (USA)
Country OLPC Brazil Argentina Thailand Libya Nigeria Palestine OLPC OLPC
Adaptor (Plug) EU US AU US EU UK EU US US

Orphaned?

Checking the Orphaned pages I found this one... which doesn't seem very logical. There are 142 other 'lost' pages, some are definitely worth losing, others, I'm in no position to tell. Thought I raised the issue. --Xavi 22:16, 2 December 2006 (EST)

Maybe this page is worth losing, since I can't even edit it. That's probably why there are so many lost pages; in general all the logical places to link from have been "protected". (this "protection" is like not being able to repair the roof on a historic building) AlbertCahalan 17:53, 17 March 2007 (EDT)
I'm no position to argue if this page should be protected or not—although it seems to harbor some hard data that could prove sensitive if edited carelesly or by mistake.
Notwithstanding, an orphan page can only be de-orphaned by editing other pages, so its protected status is irrelevant... --Xavi 18:44, 17 March 2007 (EDT)
I meant that differently. This page is difficult to usefully de-orphan because so many other pages are protected. I could de-orphan it by linking it from your talk page, but that sure wouldn't be useful. I did in fact find a semi-useful place to link it from, but there were many better places that I was unable to edit. Going the other way, I can't use this page to de-orphan something else. I could copy the page to something editable, but that probably isn't so great for the quality of this wiki. Most of the protected pages have very few links to other pages, despite obvious need. I suppose you could fix the problem by handing out admin rights to all non-spammer users, but unprotecting all the pages looks like the logical solution. AlbertCahalan 21:28, 17 March 2007 (EDT)
What I don't understand is why you don't just make suggestions on the discussion pages of the protected pages when you think there are changes that should be made? --Walter 21:41, 17 March 2007 (EDT)
In linking a wiki, overall quality comes from a great number of very tiny changes. I could indeed say "under header W, Nth sentence in the Mth paragraph (currently starting with "X"), please link the word Y to article Z. I could do that hundreds or thousands of times. That would be way too tedious for me, almost certainly way too tedious for you (but I don't know your personality), and a major source of talk page pollution. There is also something unpleasant about working hard to write complicated notes that might never get noticed. AlbertCahalan 22:21, 17 March 2007 (EDT)

Make the length byte unsigned

I am new to this, but how about making the length byte unsigned, since the 8th bit of this byte is unused.

more dates?

When is the software download date (SWDL) set? Do we need an activation date as well (for warranty purposes)?