RFRFPFilter: Help me choose a Web Developer for this huge site!

December 19, 2007 10:53 PM

I've been charged with redoing our Label Website pretty much from scratch, both back-end and front-end (yes, its a self-link but its relevant). Because this site is about classical music, there is a HUGE amount of metadata that will need to be collected, parsed and displayed (think 3700 titles * 10 tracks * performer, composer, title, length, etc for each track). Has anyone done or know of anyone who's done work on this scale on either the back-end side or the design side before that they might recommend? I have an RFP pretty much already mocked-up. I would've liked to have used someone like 37signals but it seems like they don't do work for others anymore.
posted by softlord to computers & internet (13 comments total)

Um, yes. But it's my company. We've done plenty of high-level metadata craziness, scalable stuff, clean interfaces and intelligent database design.

You'll probably see some more of us who will say the same.

Feel free to email me, especially if you have an RFP already mocked up.

And consider me slightly surprised that Sony/BMG doesn't have an in-house team for this sort of thing.
posted by disillusioned at 11:35 PM on December 19, 2007

I asked a somewhat similar question l(which the OP answered) last year. From my own explorations on the database design side, this will be complex to do correctly, but well worth doing right. I'd love to see your RFP.
posted by orthogonality at 1:05 AM on December 20, 2007

You could also try posting this to Metafilter jobs.
posted by Brandon Blatcher at 2:23 AM on December 20, 2007

You can also find qualified people on Rent-a-coder

http://www.rentacoder.com/RentACoder/default.asp
posted by DudeAsInCool at 2:31 AM on December 20, 2007

3700 titles really isn't a lot of data (I've worked on sites with serveral million records (of 30+ fields) being parsed and updated daily)

Feel free to email me your RFP. I can probably get something out to you this afternoon.
posted by missmagenta at 2:56 AM on December 20, 2007

missmagenta writes "3700 titles really isn't a lot of data"

No, it's not, but that underestimates the nature of the problem.

I'd want the database, if I looked up "God save the Queen", to show me all versions of that, including Liszt's S 235, "Variations for Piano on God save the Queen" and Beethoven's WoO. 78 "7 Variations on God Save the Queen". For all of these, I'd want to see all publications (CDs, records, tapes, etc.) by all performers for each "song".

And it should link to Haydn's "Gott erhalte Franz den Kaiser" ("God Save Emperor Franz"), and to the Prussian national anthem "Heil Dir im Siegerkranz". And to the use of the tune in Clementi's WoO. 34 Symphony No. 3 in B major ("The Great National"), which uses the tune to "God Save the King" as its theme.


If I looked up "Lili Marlene", the database should know that that title was first rendered as "Lili Marleen" but that the song was originally titled "Das Mädchen unter der Laterne". So all three titles ought to point to a single composition, and looking it up by one title ought to result in a list of all performances of that composition, regardless of the name.

But wait! It gets harder, because that composition has been recorded in (at least!) both German and English. And in both languages with the original lyrics and with revised lyrics. The database ought to have a way of distinguishing different sets of lyrics that are sung to the same music, while at the same time presenting all performances which are a "type" or "kind" or "variation" of the (Platonic ideal of the) "Lili Marlene" song.

One hopes the database will mention that the original German lyrics come form the poem "Das Lied eines jungen Soldaten auf der Wacht" by Hans Leip, the music was composed by Norbert Schultz, and the English lyrics by Tommy Connor.

But wait! It gets harder, because the music to Lili Marlene was also used, with entirely different words, as the parody "The D-Day Dodgers" (in which the British 8th Army musically replies to Lady Astor's slanders). So a look-up of "Lili Marlene" should also include a footnote or link to "The D-Day Dodgers".

What would be even more useful (both to a casual researcher and to a site selling CDs), would be to put "Lili Marlene" in its WWII context as a song (and its WWI context as a poem), to give a short biography of its best known performer, Marlene Dietrich, to link to other songs by her, and to link to other songs of war, both contemporary and historical.


Now, if I look up Phil Och's "Here's to the State of Richard Nixon", not only should it link to its precurser, Och's "Here's to the State of Mississippi", but to various covers and revisions including Eddie Vedder's "Here's to the State of George W", and ....

In other words, every performance shoudl be put into the context in the history of all performances of that composition, and its history as part of the progression of the performer's work, and into the context of the compositions that influenced the composition in question, and the compositions the composition in question influenced. The database should be able to capture all of those relationships.

I've made a few attempts to sketch this out in DDL ("database definition langugage", that is the tables that capture these relations), and (speaking as someone who understands databases rather well) it's an extremely complex problem.
posted by orthogonality at 5:18 AM on December 20, 2007

Discogs does it.
posted by mkb at 6:41 AM on December 20, 2007

I fail to see the complexity on the programming side. Collecting all that metadata, yes, that's HARD.
posted by dhoe at 7:25 AM on December 20, 2007 [1 favorite]

Enlighten did the programming for the NorthSide catalog app. It has some features not often seen, such as searching by instrument. Contact me via email in the profile if you'd like more background.
posted by omnidrew at 7:45 AM on December 20, 2007

try BuilderConsulting.com in Kansas City. They're a one-stop shop: they do lots of solid database back end work and front-end design as well as hosting and maintanence.
posted by wmeredith at 1:22 PM on December 20, 2007

orthogonality has it. Its not just the amount of data, its the different ways of slicing the data that will make it a monster. Think something like classical.com as a baseline (browsability-wise, not design-wise).

I will post to MeFiJobs and look into the parties mentioned. Thanks all.
posted by softlord at 3:24 PM on December 20, 2007

Incidentally, I am available to free-lance, and this would be a project enjoyably challenging. I should warn you, though, it's not a quick or cheap undertaking -- to do it right means managing a lot of complexity. Drop me an email if you want.
posted by orthogonality at 7:26 PM on December 20, 2007

orthogonality has it.

No, dhoe has it. This is human knowledge, not computer knowledge. If you have the information stored in some kind of database form, then this will work. If not, a million supercomputers working in parallel until the end of time will not figure it out. You need to get the data. When you have it, it hardly matters if you have 9 billion records. Once you have the data, if I had the time (I don't) I could program this for you in about a month max.

Also, I realize that self-linking kills babies, but I think that perhaps I might not understand the real scope of this website and maybe it would be useful to put a link to it somewhere in your profile? It is, effectively, going to be like Amazon.com but limited exclusively to classical music, right?

Now, if you want to create a system for entering in the data, this does make it more complex but again most of the time will be spent by around 100 otherwise out of work music theory majors punching in information into a workstation.

The website itself? Piece of cake really.
posted by Deathalicious at 2:45 PM on January 10