Thanks for the reminder that XML is still important. Since I moved into the AI field 10 years ago, my biggest battle remains convincing people that their content has to be clean and curated. Garbage in/Garbage out applies to AI. AI is not going to “fix everything” as some people expect.
It's wonderful to see XML and structured content coming full circle again with the challenges of quality AI content generation. Thanks for the call-out!
I would argue that XML on its own did not take it far enough to get us to interoperability. When I started out in the business, we had one XML standard, DocBook, and it presented its own limitations to interoperability: Few orgs I worked with left the DTD unchecked, they modified it to fit their domain model or writing rules. Other orgs created their own DTD from scratch. Both of these approaches were well suited for better consistent authoring within an org but terrible for interoperability.
DITA was supposed to break through this limitation with its novel extensibility mechanism for specialization -- but too few organizations were willing to tinker with the model and decided to use the out-of-the-box solution. Many orgs fell short on properly funding their DITA deployments, choosing not to grow headcount or invest in real training to meet the promised ROI. The real uplift DITA offered was missed when teams simply converted their content from FrameMaker and moved it into highly-capable component content management systems (CCMS) to produce more PDF output.
Orgs that did properly invest in their people to build capabilities around publishing and automation and took advantage of DITA's native strengths to deploy tight content models for more consistent content are clearly ahead of the game now on the doorstep of AI-readiness. We're at that inflection point where we'll finally see the real benefits of strategic DITA adoption.
You’re bringing back old memories. In 2000, I was leading the web UX team of Tridion CMS in Amsterdam 100% build on XML!
The big shift IMHO is when IE started to support XML local storage (we used it as caching mechanism) and .Net introduced namespaces so you can have unlimited independent XML languages I might say.
I even flitted with DITA but that one is a beast 😂
Funny how json took over and now Anthropic best practice says their LLM understanding of a user prompt increases by 20%-40% if you use XML tags (<goal>, <constraints>,…).
Thanks for the reminder that XML is still important. Since I moved into the AI field 10 years ago, my biggest battle remains convincing people that their content has to be clean and curated. Garbage in/Garbage out applies to AI. AI is not going to “fix everything” as some people expect.
It's wonderful to see XML and structured content coming full circle again with the challenges of quality AI content generation. Thanks for the call-out!
I would argue that XML on its own did not take it far enough to get us to interoperability. When I started out in the business, we had one XML standard, DocBook, and it presented its own limitations to interoperability: Few orgs I worked with left the DTD unchecked, they modified it to fit their domain model or writing rules. Other orgs created their own DTD from scratch. Both of these approaches were well suited for better consistent authoring within an org but terrible for interoperability.
DITA was supposed to break through this limitation with its novel extensibility mechanism for specialization -- but too few organizations were willing to tinker with the model and decided to use the out-of-the-box solution. Many orgs fell short on properly funding their DITA deployments, choosing not to grow headcount or invest in real training to meet the promised ROI. The real uplift DITA offered was missed when teams simply converted their content from FrameMaker and moved it into highly-capable component content management systems (CCMS) to produce more PDF output.
Orgs that did properly invest in their people to build capabilities around publishing and automation and took advantage of DITA's native strengths to deploy tight content models for more consistent content are clearly ahead of the game now on the doorstep of AI-readiness. We're at that inflection point where we'll finally see the real benefits of strategic DITA adoption.
Thank the Stars!
You’re bringing back old memories. In 2000, I was leading the web UX team of Tridion CMS in Amsterdam 100% build on XML!
The big shift IMHO is when IE started to support XML local storage (we used it as caching mechanism) and .Net introduced namespaces so you can have unlimited independent XML languages I might say.
I even flitted with DITA but that one is a beast 😂
Funny how json took over and now Anthropic best practice says their LLM understanding of a user prompt increases by 20%-40% if you use XML tags (<goal>, <constraints>,…).