Easyling release notes September, 2025

With a bit of writers’ block I was staring at a blank page for a bit when it clicked: it’s the end of September. I can make a joke about Green Day. But I must have made that last year! Luckily I checked, I managed to not do it in the past 5 years. Now with all this meta info, it wouldn’t be funny anyway so let’s jump straight(ish) into it. This month, we focused on introducing many quality-of-life improvements to Easyling. Read the full article for the details.

New supported tags in Crest

Crest, the JavaScript translation engine, is designed to support as many web pages as possible. These all pose their individual challenges. This month, we added support for localising iframe, video, audio and source tags. These are now treated as Resources. You can use the Pages list to replace them to pre-localised versions.

Wildcards for Ignore IDs

With the update this month, we introduced wildcard groups into Ignore IDs. You can now use * to mean “anything at the end”. For example, you can specify ai-* to exclude all elements where the ID starts with ai-. Note that this feature only supports prefix matches, so the star can only be at the end. *-result and ai-*-response are both invalid.

This feature, alongside the other ignore options, is particularly useful when you’re ingesting content via serving through Crest. This method gathers any content that is visible in the visitors’ web browser. As a side effect, it can ingest content from extensions. The worst tend to be those that show AI-generated content (like summaries) overlaid on top of the page. That content can be ingested for translation. Use prefix exclusion rules to prevent this.

HEAD requests vs public cache

HEAD requests are similar to GET and POST except that they are used to only transmit the headers of the given URL. Most servers handle GET and HEAD requests identically. We identified an issue where a server would send an empty response to HEAD requests and that would pollute our cache for the GET request. As a result, the translated site would not load.

We resolved this by never storing responses to HEAD requests and by using GET requests for most scenarios.

Miscellaneous

“Surprisingly”, we released many smaller fixes and features too.

1. The Workbench now checks the target language more strictly. Suppose you’re translating a project both to Spanish (Spain) and Spanish (Mexico). When translating to Spanish (Spain), the Workbench now won’t show suggestions for Spanish (Mexico).
2. We added an option to specify a custom source domain for the language selector independently of other project configurations.
3. When exporting to CSV, some segments can cause issues. For example, if you have commas in them that can break formatting. Entries where rendering fails are now skipped.
4. When using gateways to access the source site from specific IP addresses, some HEAD requests would fail. We rectified this edge case.
5. Updated XTM API stubs after WSDL changes. This ensures our integration continues to function as expected.

This document is 100% human-generated, typos and all.