Translate main website

[joyously]

It might have privacy issues, but there is no infrastructure changes, multiple copies of content to synchronize, limitations of which languages are done, time expended on translating, and it works easily even if you change the original daily.

[viktor]

That’s a good, easy option. However, we need to check and see if it is GDPR compliant. My guess is Google Translate is similar to Google Fonts, most likely logging user IP when translation is requested.

[timkaye]

I’ve just had a look, and it’s very hard to find out! But I suspect the same as you, and there’s no way we can afford to take that risk unless we are 100% sure it doesn’t exist.

[viktor]

It’s not great:

Is Google Translate Bad For Privacy & is it Secure?

[timkaye]

Yes, I found that. It’s not clear even from that, but it does rather suggest that we can’t trust GT.

[viktor]

They upload it to the server to process it, and that request must be logged so IP is logged. That’s my takeaway.

[timkaye]

Makes sense.

[ElisabettaCarrara]

Translating with a machine with no human intervention? Oh so you trust an AI with Arab, Russian or Japanese? Good luck.
The main origina site (source) holds the original content, and a bunch of files in GitHub. When source is updated the update gets committed to the GitHub AND this triggers teams to translate it and update files.
The main site doesn’t update daily. The blog posts that come out can be translated in about two hours each.

If you really like the idea of relying on an AI to translate Google is not the only option out there. But professionally I would not recommend it. It makes a site look unprofessional.

I can understand the concern about people needed to do this, but if you want to do something you can start small with just ONE localized site and add others as people volunteer (so everytime a localized sub community arises you can welcome them and entrust them with translating the main site).

[williampatton]

I understand the distrust for AI to translate the site. I don’t expect AI can do it with zero errors. I expect it can do it well enough though for a baseline.

For decisions like this I always suggest using a decision matrix to help make the appropriate path more clear. Here’s my quick decision matrix:

Use of a frontend only translation widget like Google translate

• Cost: free
• Implementation time: 1 hour
• Ongoing maintenance: minimal to zero
• Chance of translation errors: high
• Chance of user mistakes: zero
• Work required per language: 0 hours
• Languages available: many

Use of MetaCAT, multisite and manual transfer with 1 person per language

• Cost: $5/month in server costs
• Implementation time: 12-20 hours
• Ongoing maintenance: high
• Chance of translation errors: low-moderate
• Chance of user mistakes: high
• Work required per language: 1 week (estimate taken from discussions here)
• Languages available: limited to contributor knowledge and availability

Use of a multi-language plugin and their translation services

• Cost: $5/month in server costs, $20-40/month translation services costs, $99/year for multi-language plugin
• Implementation time: 6-12 hours
• Ongoing maintenance: moderate
• Chance of translation errors: moderate-high
• Chance of user mistakes: moderate
• Work required per language: less than 1 day + translation costs
• Languages available: moderate amount available

With this decision matrix the clear option to get immediate value and bring no maintenance burden is a translation widget like Google translate or other free use system.

The other 2 options are costly in terms of time required and have a directly measurable monetary cost of at least $5 what would need offset by value brought from the translations being made available.

[ElisabettaCarrara]

Your matrix doesn’t take into account the results in brand image and awareness.

Damage in the long run in terms of reputation and adoption rate:

1. High
   2 low
   3 moderate/low

That is the thing oftentimes gets overlooked.

That is why I don’t trust gtranslate and co.

But I am only one.

[williampatton]

Right, the matrix is just a starting point from my perspective and it welcomes input and details from others with a different point of view

[viktor]

Since we can’t slap a Google Translate on the website for privacy reasons, that’s a no. But, since many use Chrome they do get auto-translate. I do for certain languages I have to deal with. Most users know what to do, it’s 2022.

Once we update the main website, here’s what we can do:

1. Clone main website to a staging sub-domain.
2. Enable multisite.
3. Create a test sub-site for Italian.
4. Figure out what’s missing, what we need to add, and if it’s feasible to do at all.

Because there’s a lot of work to be done right now, this would most likely be mid to late July. There’s no rush.

This discussion will lock in 2 days, you can add a few final comments if needed. I will re-open discussion once I’m ready with multisite.

[timkaye]

Excellent plan, Viktor!