Miniature disasters and minor catastrophes

KT Tunstall’s wonderful song is playing on Pandora as I type this, and it’s just so fitting I have to use it as this post title!

This is a tale of beating DSpace and OS X with many, many rocks until they sorta-kinda work. I present it here in hopes of sparing someone else considerable annoyance.

One of my best clients emailed me with a “please fix this link in my HTML item” request. Simple enough, right?

The said HTML item is nested in folders three deep. This means that DSpace’s regular exporter breaks, because it’s not smart enough to create intermediate folders. Joy.

So I kicked that up to the dspace-tech list, and got a kind response from Larry Stone of MIT: “use the METS packager export instead.” I did, and lo! it worked.

So I twiddled the file needing twiddling, zipped up the whole, and tried to put it back. First the METS ingester barfed because I’d zipped the folder containing all the files, not the files themselves. Okay, durrr, I felt stupid and zipped the files properly.

Then the METS ingester barfed because unbeknownst to me, Mac OS X’s native zip utility adds OS X-specific junk into the zip file. Quite properly, the ingester said primly, “Your METS manifest doesn’t match your actual files. Go forth and fix it.” The solution to this little difficulty turned out to be YemuZip, which can emit a normal zip file.

Then the METS ingester barfed because the file I’d twiddled was a different size from what the METS file was claiming, logically enough. Helpfully, the ingester’s error message told me what size the file actually was, so I could pop into the METS file and fix the size in the several places it appears.

Then the METS ingester barfed because the checksums in the METS file didn’t match the checksum of the file I’d twiddled. There’s probably a quick and easy way to calculate a checksum from the command line, but CheckSumApp has a cute little GUI. Like the file size, the checksum appears several places in the METS file, so I made sure I got all of them.

Then the METS ingester actually worked. So now I have to go in and do database magic so that the item handle points to the new item, because the METS ingester doesn’t have a replace option the way the normal ingester does.

Anybody who thinks that a normal repository manager is going to go through all this to fix a link in an HTML file is as barking mad as I am. This is the ridiculousness that DSpace’s insistence on no-versioning, butterfly-pinned-to-wall “final archival” reduces me to. Yes, it’s funny—but it also cost me an entire hour to fix one link.