Marcel Klehr

How about https://github.com/sofish/pen ?

Mh. I assume this is because the URL implementation simply treats `;` as data, which is fine, but it's not canonical. The above-mentioned RFC says: > reserved = gen-delims /...

Incidentally, ```js const a = 'https://my.otrs.dom/index.pl?Action=AgentTicketZoom;TicketID=707128'; const b = 'https://my.otrs.dom/index.pl?Action=AgentTicketZoom%3BTicketID%3D707128'; new URL(a).search === new URL(b).search //=> false ```

Alright, so the [URI spec](https://tools.ietf.org/html/rfc3986) is being superseded by the [URL spec](https://url.spec.whatwg.org), which uses the application/x-www-form-urlencoded format for the query string and that doesn't seem to care about the reserved...

This is where the allowed mime types are listed: https://github.com/nextcloud/photos/blob/master/lib/AppInfo/Application.php#L36

A simpler approach might be to have admins set up [tensorflow serving](https://www.tensorflow.org/tfx/guide/serving) with a pre-trained resnet model and build a simple app that shoots prediction requests at the HTTP endpoint...

That would be awesome, of course. There's [php-ml](https://php-ml.readthedocs.io/en/latest/), but it doesn't allow loading pre-trained models, so every instance would have to train their own classifier, which is not that cool,...

> sending user data to a service for machine learning systems violates that hard The goal would be to never send user data to a third-party for classification, but have...

Just found [rubixml.com](https://rubixml.com/), which might be interesting.

@andrewdalpino Being able to efficiently run a convolutional net would certainly be a big step. Another step would be [the ability to load models trained with other frameworks](https://github.com/RubixML/ML/issues/58), which makes...