Thin Ice

Allchin: minor WMP update this year

Thomas Hawk passed on news of a minor update to Windows Media Player.

While we're both hopeful of performance improvements for larger media libraries, my main hopes are that they will first address

• overwriting of metadata, track file-names and even deletion of tracks which has been evident since v9 was in beta
• user control over album art so that correct art (like other metadata) is not overwritten by WMP when it decides your art is wrong: 1 2 3
• album updates and search 1 2
• linking of error IDs to online resources

Allchin says no WMP fixes until Longhorn

Thomas Hawk dined with Jim Allchin, and extracted the following titbits:

Microsoft is building the next version of Longhorn Media Center to accommodate much larger digital libraries. As some of you may have read in the past I have had some performance issues with my large .mp3 library. Longhorn will be better able to handle very large digital libraries. Having 500 photos or so in a folder will render quickly and the media player should be able to easily accommodate 100,000 + item media libraries.

Jim did acknowledge that there still were some issues today with Windows Media Player and I shared with him the fact that earlier that day Chris Lanier had blogged that his collection of WMP web help articles now had seen over 1 million views. Jim said that they are working on many of these issues but that a new version of WMP and fixes would not be out until Longhorn. [My emphasis].

I wonder how much user data - files and metadata - will be destroyed by WMP by then? Will the Windowsmedia team actually act constructively regarding user feedback on bugs that they have ignored through two previous betas? Does this translate to fixes for Windows XP and long overdue fixes for Windows 2000 versions of Media Player?

Online store opens in New Zealand

Six months ago, I changed my Windows User Input Locale setting from English (Australia) to English(New Zealand) to evade the ghastly online stores that had invaded my desktop. When these stores turned up, the album reviews and track listings available through the windowsmedia.com database (the *only* database accessible to WMP users: it's not CDDB and it's not user configurable) were replaced by some really poorly presented rows of unrelated albums generated from the od2.com fugly thumbnail catalog that the vendor wanted to sell. Over the last 6 months, I can confirm that the presentation of this catalog hasn't become any more professional. The ninemsn suggestions are still laid out in a stupefyingly primitive manner.

Tonight I noticed that Digirama had opened as the NZ Music Store, and I thought I may have to reset my locale to English(Belize) (which at time of writing, was still spared the incursion of online stores). Fortunately, the View Album Info function has not been relinquished to the the OD2 fugly thumbnail server and you can still get album reviews from the windowsmedia.com.

If that is lost, I guess I'll be grabbing my scuba gear, sand-fly ointment and heading for Belize...

Bundling of Proofing Tools

Michael Kaplan blogged on this subject.

It's very true that the costs of producing any of the language tools can vary enormously from one language to the next. In general, I expect that the smaller the language-base, the greater the cost. This is due to there being less choice resources available to choose from in terms of say lexical resources, voice or handwriting samples and computational linguists. Frankly, if you have a population under about 10 million, how many people do you really have devoted to collecting and analyzing the necessary data for creation of these tools?

[Digression]
In smaller or less-developed countries, the pattern is for interested parties: government, educational and media organizations to pool such resources to create a standard set of spell-checking, thesaurus or possibly voice-recognition tools. They can then tailor them to different APIs for Microsoft Office, OpenOffice or other platforms.

There are also some specialist companies who deal with the unique issues of low population language bases or even creole populations. I am sure that some of these are done as a labour of love by modern-day Samuel Johnsons.

For the record, there is not just a single spell-checker for Office and multiple "dictionaries" for each language. Each language has a unique spell-checker (and grammar-checker and thesaurus and hyphenation and ...) DLL that comes paired with equally unique "LEX" files which store linguistic data. There is not even a single "LEX" format: it's really just a Microsoft convention for naming the resource files that accompany the checker-engine DLLs. And since Microsoft ships DLL/LEX sets produced by many other companies, the LEX format will vary according to the imagination and computational skills of each of these companies.
[/Digression]

Nonetheless there are definitely issues with making even the Proofing Tool sets available to the market. Each version of Microsoft Office sees a new version of the PT CD released. The versions of Office and the PT CD must match at installation time but you can subsequently upgrade Office and the older PT elements will still operate. Unfortunately, older versions of the PT CD generally disappear from warehouses long before their product lifetime has expired, and even the Microsoft online store struggles to keep them in stock. For several years I had to send an email to a senior member of the Microsoft Word team to ask them to prompt distributors to keep these items in stock. I hope that Microsoft can figure a way to make these tools available over the web.

How much worse will Windowsmedia online database get?

I've remarked before on a case of Windowsmedia reporting different album information according to whether or not you click the "Find Album Info" or "View Album Info" buttons in Windows Media Player v10. Today I picked up a copy of the latest Verve remixes, and threw it into the computer to rip to Lossless WMA.

WMP didn't automatically pick up the album details, but when I clicked the Find Album Info button, it showed the correct album "Various Artists - Verve Remixed, Vol.3" ... but no track information was included (or album art). Which makes me wonder why AMG even bothers to put the information into the database. I ran a search on "Verve Remixed" but did not find any other records for this, Volume 3. So I returned to the original record, edited the tracks manually and uploaded the details.

Out of curiosity, I clicked the "Find Album Info" - this time I get a different record returned which has all the track titles (but still no album art). Note, these are NOT the ones I uploaded, as I included the remix information in the titles.

It makes me wonder further, if Microsoft and AMG et.al. collectively can't give a damn about maintaining these databases or getting the search code to work, why WMP is not opened up to allow access to other information sources. Since WMP is an integral part of Windows (as we hear so often and so strenuously from Microsoft), doesn't the tying of album-links to a single database basically amount to the same thing as smart-tags redirecting one to a proprietary Microsoft back-end?

WMP Year/Date Released

There are always new unexplored corners of Windows Media Player ready for experimentation. Usually when I'm in the Library pane, I live in one of the Artist or Composer views. This evening, I expanded the Year Released tree to see what was listed.

First up, you see that the Year Released tree is based on the Date Released field, another of those "let's give every field multiple names" cases in WMP. (See the article on Artist fields on this site, for more examples.)

The expanded nodes start with "Unknown" (ie no data in the field) and then range sparsely from 1899 to 2019. As noted in the WMP Library bug-list, WMP likes to add or subtract 1 from some year ranges, so "1899" in the Year Releasednode corresponds to "1900" in the Date Released tree. This bug has been around since before v9 was released. It's also noticeable in the way that Windowsmedia/AMG album descriptions are usually one year different to the year that is downloaded as track metadata.

[2006 update] Apparently this bug affects all users whose systems operate forward of GMT timezone i.e. GMT+X. Will there be a fix, or will that wait until the US somehow moves into these timezones...?

---

The few nodes with years > 2005 just have a random selection of tracks that have been mis-labelled. I discover to my surprise that I cannot erase the field (and thereby shuffle these tracks into the Unknown node) or drop the tracks directly onto the Unknown node. All I'm left with is changing the field to another number, so I bulk edit the tracks to be "1900" which um...naturally drops them into the 1899 node. The Advanced Tag Editor does not give you access to this field, so that route is also closed off. I'll probably have to edit the tracks in another program to clear this field.

So, while the layout of trees is visually similar for Album Artist, Contributing Artist, through to Purchased Music there is a fair amount of variation in the way they actually behave. Some are single-valued fields (e.g. Album Artist, Album) where a semi-colon delimiter cannot be used to group a track under multiple headings. Others like Composer, Contributing Artist and Genre are multi-valued: the semi-colon delimiter can be used in this fashion. Within the latter group, there is the further difference that drag-and-drop works differently for Genre. If you drag a track to a Genre node, you are given the option to Change or Add Genres. I find it peculiar that this option does not exist for the other multi-valued fields. And then of course we have the single-valued Year Released which does not allow you to drag tracks onto the Unknown node. This restriction does not exist for the other fields discussed. I *really* hope that none of the folks who designed this have anything to do with WinFS or other metadata-handling elsewhere in Windows UI.

I'll mention in passing that there is also a "Period" field that accepts text, so you could write something like "Baroque" or "1940s". This is somewhat redundant given the multi-valued Genre field already allows me to create such labels. If the "Period" field were coupled with a left-hand pane tree view it might be more useful, so that chronological categories weren't lost amongst the thematic Genre items.