The articles that I send to my Kobo are a big part of my information diet.

This workflow broke when Mozilla announced it was shutting down Pocket. For the past seven months I have been trying to find a replacement, and migrate all my data.

Long ago I had used Wallabag, and found the interface clunky, and the article extractor inadequate.

Eventually, I settled on the easy option: just use Instapaper, and which Kobo had chosen as Pocket’s successor.

It’s been around a long time (I used it before Pocket), so it was likely to stay around for a while, which was suddenly a very important criteria.

Convenience (good Kobo integration, slickness) won out over principles (Instapaper is proprietary).

Until it didn’t.

When I imported my archive into Instapaper, 1,000 articles silently disappeared.

The official Pocket CSV “export” didn’t even include starred status of articles. Fortunately, I heard about this in time, and had self hosted and run https://github.com/ArchiveBox/pocket-exporter, so I had a complete export of my Pocket data, and had already explored it, so I knew how many articles should be in Instapaper.

I wrote back and forth with Instapaper support, and they were very helpful, and eventually they had me re-import from Pocket CSV (previously it had pulled directly from Pocket API, before Pocket shut down), and all the entries were there.

But now all my favorites were gone! While I was waiting for Instapaper to sort out the missing articles, I had gone ahead and imported everything into Wallabag, and discovered that my favorites were gone. But Wallabag has a good API, so I wrote a little Python script and jq pipeline to get all my favorites from the more complete, unoffical Pocket export, and star them in Wallabag. It worked great.

Since Instapaper couldn’t recover the starred status on their end, I requested API keys so I could adapt my star_wallabag_article_by_url.py script for Instapaper.

I adapted my Wallabag script, and they granted my API key request.

This is when I discovered that the “full” Instapaper API has some serious limitations. Officially, it has an "optional" limit parameter, which should be a number between 1 and 500. Unofficially, it never returns anything after 500 entries, even with the correct offset/have parameter.

I’m not the first person to discover this.

I contacted Instapaper support again, like the author of this blogpost, and never heard back. To be fair, I haven’t paid them money, so they don’t owe me a response, but I feel that they should make this limitation clear.

If Instapaper isn’t more convenient than the open-source competition, there’s no reason for me to use it. So I researched Wallabag instances, and installed KOReader – and KFMon and Plato – on my main e-reader. Previously, I’ve relegated KOreader to my spare, junker e-readers.

This is when I encountered bitter medicine:

1. The Wallabag.it article extractor failed for 1,108 of 3,947 articles. In practice, this is not super relevant, because ~608 were still unread in Pocket, which means they failed to extract and send to Kobo. Of the remaining ~500, I saw a lot of dead URLs (feedproxy.google.com et cetera), which aren’t Wallabag’s fault. I also discovered that while adding a URL directly fails for some sites that worked with Pocket (such as www.mrmoneymustache.com), the Wallabagger extension can extract their content no problem (or, you can save an archive.org snapshot into Wallabag)

2. Large Wallabag v2 JSON imports are unreliable on most instances. I have yet to find any other Wallabag instance that can handle importing my full Wallabag.it export without timing out or giving a 500 or 413 error.

3. You can’t star Wallabag entries from KOReader. What?! Fortunately, it was simple to patch KOReader to support this vital-to-me feature.

In summary:

Wallabag is slow, the Kobo integration is inferior in the ways that matter to me, the article extractor needs work, imports are a headache, and most instances aren’t kept up to date, and this volunteer-run project is still better than the well funded competition!

It’s better because it can import from all kinds of services, it runs on practically any e-reader, it’s open source so it can’t be shut down as easily, you can highlight your articles, you can hack anything about it to be exactly the way you want it instead of just complaining to customer support, there’s a huge ecosystem, it’s been around for ages, it has a powerful and easy-to-use API that means you rarely need to use the web dashboard, and it supports full-text-search.

Wallabag wins, despite everything.