Once #13 is merged in, email ID is consistently available regardless of whether user is whitelisted or not. This is a great time to implement exponential back-off strategy that would ensure that email gets processed until API responds with message ID.

cc @alastairong @jarodholtz

I noticed that the following part of challenge template:

<a href="https://{{holochain_agent_id}}.holohost.net">{{holochain_agent_id}}.holohost.net</a>

Has been changed to:

<p>https://{{holochain_agent_id}}.holohost.net</p>

I'm not sure what the motivation is. I'd guess that it was made into text rather than a link because we previously had "Link not working?" text there, and that no longer applies, it's a completely different link to the one that is used as part of the button.

Additionally, I optimized the template so that the entire address would fit, and because https:// is now visible part of this line, it no longer fits:

Unless anyone minds otherwise, can I revert this back? If no, I'll just close this because this is way too minor to have an argument over.

If IsWhiteListed check returns something falsey we should send an “Invalid Email” email for easier troubleshooting.

Template text found here: https://hackmd.io/_D_JAX7qS9a14GVbE-3vtA?edit

This should be implemented before deploying holo-auth to production.

A HoloPort owner was having trouble registering his email address because he started it with a capital letter, whereas the whitelist had it in all lower-case. I see that the worker doesn't do any normalisation for upper/lower case, and indeed it can only normalise the email address in the user's registration request, not the whitelisted email address.

Most email servers don't enforce case-sensitive addresses, so [email protected] should be functionally identical to [email protected] for 98% of servers on the planet. I feel that we should offer that same consistency.

But I'm not sure what to recommend, because you can't normalise the whitelisted email addresses -- it's a key in the K/V store, after all. We could require C/S staff to always input email addresses in lower-case; that might help, but it's error-prone.

CC @yegortimoshenko @alastairong

cc @alastairong

Limiting how many times email can be used to activate HoloPort is currently not implemented, as discussed previously. Filing this issue to reflect that fact.

We can easily distinguish between the two based on the error fetch API returns. This will provide some more context into the init process.

During client Rust rewrite, I forgot to add in ZeroTier status check:

https://github.com/Holo-Host/holo-nixpkgs/blob/b33dc9a4718ed7d1a5859b6b5eb3f7289b6cbb82/overlays/holo-nixpkgs/holo-auth-client/holo-auth-client.py#L54-L57

It should be added back in.

Whitelist KV store is currently empty. It needs to be populated with data from Freshdesk. Write an import tool, run it, and commit it to this repo.

CS has found the host ID information that we use to track “used” email address is very useful in their issue triaging efforts. Having a database that links host IDs to emails also provides options for future membership management / security / etc.

They’ve therefore requested this feature to be added before release so there aren’t any gaps in the database.

Once #13 is merged, ID will always be available in POST /v1/challenge response, regardless of whether email is whitelisted or not. It would be useful to log it for support troubleshooting purposes.

We can't read this data from Freshdesk anymore because their API rate limit is extremely low and Paul's holoport-authorizer was regularly hitting it, which was one of the things that were causing reliability issues. In addition to that, Freshdesk is a support system, and was not created for the purpose of being a source of truth for any data.

Let's decide on UI to offer support to add/remove whitelist entries. I'd recommend to start with updating whitelist KV store directly. Here's what it would take to add a new whitelist entry:

1. Open holo-auth-server-whitelist KV store.
2. Input email into "Key" field.
   
3. Press "Add entry".

To remove a whitelist entry:

1. Choose an entry to remove.
2. Press "..." icon.
3. Click "Delete".
   

cc @alastairong

Currently Client makes a call to Holo-Auth Server in a happy mode. Accounting for edge cases will improve reliability of the service.

Possible failures (networking timeout, Postmart failure) could result in a retry.

cc @jmday

Given data that is available in various APIs, it wouldn't be challenging to create a page that would show generalized stats like how many emails we've sent out (w/o counting in @holo.host registrations). If desirable, that could also be split by countries, or any other way you see fit.

This might not be urgent, but probably makes sense to discuss, especially considering that from what I gather I broke your statistics generation tools by switching to Postmark. Sorry about that.

It would be great to have meaningful name/description set to individual ZeroTier devices. #19 maps agent IDs to names and emails to descriptions, and once it's merged, this will happen on its own for new registrations.

We should import this data from Freshdesk to ZeroTier Central so that we would also keep track of old registrations the same way.

Github Actions are set up in such a way that changes merged into main and develop branch are auto-deployed to alphaNet and devNet respectively. For a minimum protection of our dev and production environment those two branches should be protected

Whenever user is registering holoport with holochain_agent_id that has already been registered a new entry is created in ZT Central registry.

ZT Central is using address field as a unique key, but many services downstream use name field (which is populated with holochain_agent_id) as a unique key. Because process mentioned above creates multiple entries with the same name value the structure of database from the perspective of downstream services is corrupted.

So in the name of brotherhood with downstream consumers I am suggesting deleting all the old entries where name == holochain_agent_id while creating a new entry.

This would require making 1 more call after this line of a server code.