Publishing with Silicon

I’ve made shoes for everyone, even you, while I still go barefoot
Bob Dylan

I am still pinching myself to see if I am awake as I look at the whitepaper on our web site. I have worked as the tireless scribe of other companies for so long, working on 10+ whitepapers for Adobe and others in the early 2000s; it is refreshing to get to the point where we can have a whitepaper of our own. By John Parsons, no less.

Pleased as I am with John’s work, I am more happy still about the topic of the whitepaper, and the state of the technology it describes. FXG is a joy to work with, now that we are past the struggles with the beta forms of the associated products, and on to real implementations.

Web to Print Workflow

The Scene7 platform was built really well, and almost anything once can imagine in Web to Print can be accomplished between Scene7 and Flash. It will no doubt evolve, but ever since FXG 2.0 we have a very practical strong foundation.

I am not a patient person at all. It was a good year and a half of naively thinking that FXG 2.0 was right around the corner: I am the sucker that believes the projected delivery dates of software products, just as I naively believed in XML rendition standards in the late 90s and early 2000s. But better late than never, and of course there’s an explanation: it seems that Flex 4 was more ambitious than anticipated, and it got late enough that CS5 integration made sense, and meanwhile S7 was waiting for the standard to be complete for the sake of interoperability.

Adobe Scene7 FXG Web to Print Workflow

We are working on several Silicon Designer implementations on top of Scene7 now; I think it has just eclipsed InDesign Server in popularity, but the two are not mutually exclusive and our latest solution uses both. XMPie and Scene7 can interact as well: the beauty of more recent technology is that with Web Services and the various techniques of translating the different forms of XML (semantic, transformation, and rendition) interoperability has gotten much less opaque than it once was in this industry.

Silicon Designer

I like John’s observation that WTP was born in the dot com era: it has suffered from the same brittle, rush-job poor foundations of such an era until now, when Adobe finally comes to their senses and acquires a company like Scene7 with true vision, and has the guts to empower them to build something of this caliber.

We worked with Adobe a bit on their Scene7 product, and I have to say that it is some of the most promising technology out there for Web to Print. There are two big gotchas that I hope are overcome soon:

1. The text that is possible with Flash 10 is not fully functional: this stands to improve once FXG 2.0 is available, the hope is that FXG 2.0 will be fully supported. As of now the text is more like FXG 1+, it isn’t quite robust enough for our typical clients.
2. The pricing model is crazy. I think they priced it so high that they would make sure not to get slammed with too many initial implementations. $50K/year as a base price with multiple forms of transaction/bandwidth costs on top of that is hardly a SaaS model. You either pay as you go or you pay up front, they can’t ask for both…

Also, it appears in their early concepts of how the app would be used, they imagined one would hit the server for the Flash renditions! I think the whole beauty of sharing the XML model between PDF and Flash is leveraging what the client can do…

Demonstration of the Scene7 web-to-print solution

Anyway, in all my years of working with great programmers, including many at Adobe, I have never seen a group as great as those working on Scene7 web to print. I am very optimistic about its future.

The fundamental beauty of the Scene7 model for web to print is that it uses the same XML to describe the web document and the print document. It also extends the XML used in Flash (FXG) to support requirements of print such as CMYK color. Tricky, as this would ideally not be done in a separate namespace, but would be part of the core FXG spec itself. In general it is awkward how the different groups at Adobe work together: they are all focused on their own short-term deliverables and can’t often reconcile or coordinate the overlapping parts of their efforts.

Which brings us to… InDesign Server. One might have guessed that Scene7 would use InDesign Server rather than build their own form of PDF generation totally independent, with a different text engine (common with Illustrator/Flash, not InDesign) and different XML model (FXG vs. IDML). Sadly, the InDesign project does represent the ultimate in text engines, the ultimate in document feature sets for long documents, etc., but there has not been a desire to use it from the Scene7 group. They didn’t find it much of a true server product, apparently, which is quite understandable. The “server” dimension of IDS is minimalist, it is essentially the rendition half of the desktop product with a few hooks and enough “build it yourself” aspects that solution providers like us have a fairly endless stream of opportunity.

So Scene7 will hopefully become a big part of our work next year, assuming the 2 issues above are handled, yet InDesign Server will remain, especially for longer documents and those cases where extending a desktop InDesign workflow to the server is easier when avoiding issues around reconciling text and layout engines. We don’t really mind two systems, but some day I’m sure we’ll hit a hybrid case where we use both, and in some long-term road map (CS7?) they should actually get reconciled.

Adobe product managers have managed to calm down my early complaints about non-reconciliation of these two engines. One pointed out the incredible backwards compatibility responsibilities of IDS: they can’t just start over… one tiny bug in one tiny dot release can screw up a million documents for a client, they are not as agile as a SaaS shop.

In terms of SaaS; as of now, Scene7 is almost only SaaS and IDS is almost only self-hosted. It is likely that both products will cross over the other direction. We can host IDS in an EC2 environment just fine, with great scalability, yet the licensing is not SaaS friendly. In similar fashion, Scene7 can install just fine as a self-hosted software, yet they only allow this in “special” situations and tend to push for SaaS at all costs.

I noticed that the draft FXG 2.0 Specification is finally online. It appears that this will be the form of FXG implemented in CS5.

I have been interested in, and somewhat connected to, Adobe’s approach to XML for quite some time. In the mid 1990s, FrameMaker supported SGML prior to the birth of XML. In 2000, Silicon Publishing worked with Adobe in publicizing FrameMaker 6.5 as an XML-capable tool, though FrameMaker+SGML only worked with XML in a very cumbersome, awkward way.

I will never forget our first project for Adobe, which was one of the very first Silicon Publishing projects. One Friday in 2000 I went to meet Doug Yagaloff, the publishing genius that led Caxton, and he gave me a copy of Frame+SGML and said I just had to do one simple thing. Import an XML document, and export it back out. That was a long weekend! I felt like I must be very stupid, it took me forever to get anywhere at all. Thankfully, Sunday night I found a “quick guide” online which exemplifies the great patience that is generally characteristic of those working with SGML and document-centric XML. I was able to show Doug an example the following Monday – “it’s hard, isn’t it?” he smiled.

We worked with Adobe on the FrameMaker 7.0 release, which dramatically improved the XML support. Later we put DITA support into FrameMaker for Adobe, which now gives a real head start when working with document-centric XML. I am a strong believer in DITA.

That is the core, semantic XML that SGML was oriented around at its foundation. InDesign got some bare-bones support for semantic XML with 2.0, but it goes nowhere as deep as the support of real XML authoring tools. Probably more interesting in terms of Adobe technology (they bought FrameMaker but it stands outside of their main product offerings) is rendition XML, and here was an area more exciting to us at Silicon Publishing.

Rendition XML

In 1998, when I was still at Bertelsmann, one of our former employees who had moved on to Adobe told me about a very exciting new XML specification: PGML. This made great sense to me, and I was an early enthusiast. It was not long before the PGML effort was subsumed under SVG, and Adobe was a major participant the SVG spec development effort, with their representative, Jon Ferraiolo, serving as lead editor of the spec itself. The Adobe SVG Viewer became the primary way SVG was viewed on the web, while tools like Batik evolved steadily and browsers (with one huge notable exception) gradually evolved support for it. Adobe Illustrator supported and still supports SVG round trip, while InDesign offered SVG export but has since deprecated it.

On another front, rendition of documents, Adobe also participated in the most significant standard: XSL-FO. Here was a document description language highly similar to FrameMaker’s MIF, and again an Adobe expert, Stephen Deach, led the specification definition. FrameMaker never directly supported XSL-FO, but a short-lived server application, Adobe Document Server, offered XSL-FO support via its underlying FrameMaker engine. This was a great XSL-FO implementation, actually, but was not well supported by Adobe and it is now extinct.

On the surface you could consider Adobe a leader in standards-based XML for graphic and document formats. However, as I discussed earlier, there is an interesting mix of motives in the involvement of such companies in web and XML standards. When Macromedia Flash was a competitor, an “open standard” like SVG made sense, but after the Macromedia acquisition, it made less sense.

Adobe has gone down the path of proprietary XML namespaces, not unlike their competitor Microsoft. And like Microsoft, whose XAML is highly derivative of SVG, they have not found a reason to re-invent the wheel.

Three XML Namespaces

There are three XML namespaces that appear critical to the future of document and graphic description at Adobe. These are IDML (InDesign Markup Language), FLA (formerly XFL, the description of Flash, and FXG (the graphic model supported by Flex 4 and central to the designer/developer workflow of Flash Catalyst): FLA handles the complete, interactive Flash model (literally replacing the binary .fla format) while FXG is more about static graphic representation. Theoretically, FXG is open source, as is the Flex SDK, but these remain extremely Adobe-centric efforts.

FXG and FLA have some strong similarities to SVG. In fact Adobe acknowledges the partially derivative nature in the specs. Of course there are differences between what was specified in SVG and what is natural to the graphic model underlying Flash; it appears that SVG would have been difficult to implement across the board, given how Flash was built and the goals of Flex yet they used SVG tags directly where it did fit the Flash model.

FXG is becoming a very powerful specification, now that the Text Layout Framework is built into it. Flash is able to render FXG and Illustrator is able to import/export FXG. With CS5 the designer/developer workflows and the general interaction between print-centric and web-centric work should become much better.

IDML is not derivative from XSL-FO. It represents a very general similarity, especially compared to the earlier INX XML format for InDesign: it is at least a complete document object model. INX was merely instructions to the scripting DOM as to how to create the document. It is too bad that Adobe has not managed to reconcile the text engine of InDesign with that of Flash: it appears that IDML will for the near term stay quite separate from the other Adobe namespaces.

To me, FLA is the most exciting new XML namespace coming from Adobe, but it won’t really be exciting until we have an FLA server that can compile FLA to SWF quickly. Dynamic content is possible with Flash in many ways already, but the possibility of making the entire SWF dynamic and manipulating that content in arbitrary ways with XML tools should bring the form of publishing power we envisioned with SVG to life once and for all.

As they have tended to miss the boat on any server application of their technology, Adobe appears to be slow to perceive the value of such a thing (I once asked Kevin Lynch for a Photoshop server – he questioned whether anyone would want it, citing experience with Macromedia Generator). It is an interesting question which group an XFL server may come out of; such a product could be conceived as natural to InDesign Server, to Flash Media Server (or some other work of the Flash Platform group), or to Scene7 (which has very powerful SaaS rendition capabilities, some of which are based on FXG). We are lobbying…

Adobe has finally built some XML foundation under their rendition models, and we are able to attain many of the things we dreamed of back in the SVG/XSL-FO days, via XML if not via open XML standards. I don’t have big hopes for Adobe integrating semantic XML in their core products (FrameMaker being a black sheep outlier), beyond simple metadata (XMP is good enough here, but document-level metadata is trivial compared to true semantic XML). Hopefully the power of their rendition technology with its new XML underpinnings (and consequent greater extensibility) will provide a foundation that enables other companies and open source efforts to make tools that bring the deeper vision of XML publishing to life.