- From: Thad Guidry <thadguidry@gmail.com>
- Date: Wed, 6 Nov 2019 08:34:17 -0600
- To: Antonin Delpeuch <antonin@delpeuch.eu>
- Cc: public-reconciliation@w3.org
- Message-ID: <CAChbWaPEfdVG=K2irYvHBJPmCkJKPK8Oa7XySYxWyjLTJtSJ6w@mail.gmail.com>
Makes perfect sense, Antonin. I think those 3 levels are fine. A register/signup link is also a good idea. +1 Thad https://www.linkedin.com/in/thadguidry/ On Wed, Nov 6, 2019 at 8:09 AM Antonin Delpeuch <antonin@delpeuch.eu> wrote: > Hi all, > > I thought about this a bit more, and my intuition is that it would be good > to specify how services can announce that they support or require an API > key. > > I imagine we could add that in the manifest somewhere. That would make it > possible for UIs to prompt the user for an API key when applicable. > > I can imagine multiple use cases: > > - endpoints which always require an API key > > - endpoints where an API key unlocks particular features (some particular > properties or types which cannot be used by default, for instance) > > - endpoints where supplying an API speeds up the reconciliation > > So perhaps we could have three levels: not required, optional and required? > > There are other things we could consider adding, such as a user-friendly > name for the API key with the terminology that the service uses in other > contexts ("token", "password", …) which could be displayed to the user when > prompting for the API key (so that they understand better what to enter). > And/or a link to a page which explains how to get a token, for instance. > > What do you think about that? > > Antonin > On 25/09/2019 08:57, Antonin Delpeuch wrote: > > I found this old thread where the feature was discussed more extensively: > > > https://groups.google.com/forum/#!searchin/openrefine/reconciliation$20API|sort:date/openrefine/hbhT7CFy5PE/PxApScXQL0gJ > > So it looks like a GET parameter is indeed the preferred way to go, with > "api_key" as parameter name. > > Antonin > On 24/07/2019 10:25, Alan Buxton wrote: > > Great that we are all on the same page. > > > > My vote would be similar to what you outline below, but just to call the > parameter “api_token” to make it crystal clear what it is to anyone who is > looking at the URL. > > > > *From: *Thad Guidry <thadguidry@gmail.com> <thadguidry@gmail.com> > *Date: *Tuesday, 23 July 2019 at 17:59 > *To: *Antonin Delpeuch <antonin@delpeuch.eu> <antonin@delpeuch.eu> > *Cc: *"public-reconciliation@w3.org" <public-reconciliation@w3.org> > <public-reconciliation@w3.org> <public-reconciliation@w3.org> > *Subject: *Re: Feature request - Optional API keys > > > > Please for the love of OpenRefine... > > > > Let's do it like the big boys do and use an authorization token (a > parameter). Debugging bad requests are easier on the consumer. > > Bolting it together for the publisher of the service is trivial and > marginally better than headers. And easier for them to debug as well! > > > > For those publishers that later want to enable Oauth dancing, it will be > easier to support that upgrade path. > > > > <script type="text/javascript" src="https://maps.googleapis.com/maps/api/js?key=*YOUR_API_KEY*&*libraries=places*"></script> > > > > Thad > > https://www.linkedin.com/in/thadguidry/ > > > > > > On Tue, Jul 23, 2019 at 9:01 AM Antonin Delpeuch <antonin@delpeuch.eu> > wrote: > > Thank you Alan for chiming in! > > I agree with you and Owen that it would be a useful thing to support. I do > believe this should be specified as part of the API, so that it can be > supported in a uniform way. Let us keep this on our list of things to add > to the specifications. > > The corresponding OpenRefine issue is here: > > https://github.com/OpenRefine/OpenRefine/issues/2015 > > A pull request for this has been submitted in the past: > https://github.com/OpenRefine/OpenRefine/pull/758 > > Antonin > > On 7/23/19 8:29 AM, Owen Stephens wrote: > > As a member of the OpenRefine community and a contributor to the > development, I would say that we should support API keys > > > > From the perspective of this group I guess the wider question is whether > we should have a standardised way of supporting API keys for reconciliation > services - should this become part of the API? Should it just be an > optional request Header? Or just something added to the URL? > > > > Owen > > > > Owen Stephens > > Owen Stephens Consulting > > Web: http://www.ostephens.com > Email: owen@ostephens.com <owen@ostephens.com> > Telephone: 0121 288 6936 > > > > On 22 Jul 2019, at 16:09, Alan Buxton <alan.buxton@opencorporates.com> > wrote: > > > > Hi all > > > > I’m writing from OpenCorporates who provide a reconciliation service > (among other API services). See > https://reconciliation-api.github.io/testbench/. > > > > We provide all our APIs under either an entirely open basis or via an API > key. We issue API keys free of charge for public benefit users, and on a > paid basis for corporate users. We have 3 types of users: users who use our > services without an API key, public benefit users (e.g. journalists) who > use our services with a free API key and those who use our services with a > paid API key. > > > > OpenRefine does not currently support API keys, so OpenRefine users are > only able to use our API through the public endpoint. To improve > performance, make our service more maintainable and generally manage it > better, we are going to be retiring some of the underlying functionality in > our main app that supports use without API key. But we want to make sure we > can still provide this service to public benefit users, so we need to be > able to support (optional) API keys. > > > > What is the view in the community about OpenRefine providing support for > (optional) API keys? > > > > Alan Buxton > CTO > > -- > > OpenCorporates :: The Open Database of the Corporate World > https://opencorporates.com > Blog: http://blog.opencorporates.com > Twitter: http://twitter.com/OpenCorporates > OpenCorporates Ltd is a company dedicated to improving and publishing > public data under an open licence that allows and encourages reuse, > including commercially. Registered in England, number 07444723. > > > >
Received on Wednesday, 6 November 2019 14:34:34 UTC