|13 History and acknowledgments|
The name "Ipe" is older than the program itself. When I made figures for my papers using Idraw in 1992, I was annoyed that I had to store two versions of each figure: one with Latex text, one with Postscript information. I came up with a file format that I called "Ipe", for "Integrated Picture Environment", and which was at the same time legal Latex source code and a legal Postscript file.
When I wrote the first version of Ipe at Utrecht University in the summer of 1993, it created this file format directly, and inherited the name. The first versions of Ipe (Ipe 2.0 up to 4.1) were based on my experiences with Idraw, XFig, and Jean-Pierre Merlet's JPDraw, used IRIS-GL and Mark Overmars' FORMS library, and run on SGI workstations only.
Due to popular demand, I spent two weeks in the summer of 1994 to teach myself Motif and to rewrite Ipe to run under the X window system. Unfortunately, two weeks were really not enough, and the 1994 X-version of Ipe was somewhat of a hack. I didn't have time to port the code that displayed bitmaps on the screen, it crashed on both monochrome and truecolor (24-bit) displays, and was in general quite unmaintainable.
These versions of Ipe were supported by the Netherlands' Organization for Scientific Research (NWO), and I would never have started working on it without Geert-Jan Giezeman's PLAGEO library. For testing, support, and inspiration in that original period, I'm grateful to Mark de Berg, Maarten Pennings, Jules Vleugels, Vincenzo Ferrucci, and Anil Rao. Many students of the department at Utrecht University served as alpha-testers (who apparently referred to Ipe as "the cute little core-dumper").
I gave a presentation about Ipe at the Dagstuhl Workshop on Computational Geometry in 1995, and made a poster presentation at the ACM Symposium on Computational Geometry in Vancouver in the same year. Both served to create a small but faithful community of Ipe addicts within the Computational Geometry community.
Ipe proved itself invaluable to me over the years, especially when we used it to make all the illustrations in our book "Computational Geometry: Algorithms and Applications" (Springer 1997, with Mark de Berg, Marc van Kreveld, and Mark Overmars). Nevertheless, the problems were undeniable: It was hard to compile Ipe on other C++ compilers and it only worked on 8-bit displays. It was only due to the efforts of Ipe fans such as Tycho Strijk, Robert-Paul Berretty, Alexander Wolff, and Sariel Har-Peled that the 1994 version of Ipe continued to be used until 2003.
I was teaching myself C++ while writing the first version of Ipe, and it showed—Ipe 5 was full of elementary object-oriented design mistakes. When teaching C++ to second-year students at Postech in 1996 I started to think about a clean rewrite of Ipe. My first notes on such a rewrite stem from evenings spent at a hotel in Machida, close to IBM Tokyo in July 1996 (the idea at that time was to embed Ipe into Emacs!). It proved impossible, though, to do a full rewrite next to teaching and research, and nothing really happened until the Dagstuhl Workshop on Computational Geometry in 2001, where Christian Knauer explained to me how to use Pdflatex to create presentations. I realized that PDF was ideally suited for a new version of Ipe.
Ipe 5 figures were at the same time Latex and Postscript files, and required special handling to be included into Latex documents, which sometimes required a bit of explaining when talking to co-authors or publishers. While editing a figure, Ipe 5 kept a Ghostscript window open that would show what the figure looked like after processing by Latex.
Several developments that had happened between 1993 and 2001 allowed me to use a completely new approach: First, Hàn Thê Thàn's Pdflatex takes Latex source and directly produces a PDF file with a PDF representation of the text and all necessary fonts. Second, Derek Noonburg's Xpdf contained an open-source PDF parser that I could use to parse this PDF representation and to extract the processed text and fonts. Third, all relevant Latex fonts are now available as scalable Type1 fonts, and so it is possible to embed Latex text and formulas in figures that may still need to be scaled later. Finally, the Ghostscript window was no longer necessary as Ipe could use the beautiful Freetype library to directly display the text on-screen as it will appear on paper.
Directly after the Dagstuhl workshop I implemented a proof-of-concept: I defined the Ipe XML format (there was no question that Ipe 6 would have to be able to communicate in XML, of course), wrote "ipe5toxml" (reusing my old Ipe parsing code) and a program that runs Pdflatex, parses its PDF output, extracts text objects and font data, and creates a PDF file for the whole Ipe figure.
All that remained to be done was to rewrite the user interface. Mark de Berg and the TU Eindhoven made it possible for me to take some time off from teaching and research. The final design changes were made during the Second McGill-INRIA Workshop on Computational Geometry in Computer Graphics at McGill's Bellairs Research Institute in February 2003, and much inspiration was due to the atmosphere at the workshop and the magnificient cooking by Gwen, Bellair's chef. An early preview of Ipe 6.0 was "formally" released at the Dagstuhl Workshop on Computational Geometry in March 2003, to celebrate the Dagstuhl influence on Ipe.
Other than the file format, there weren't really that many changes to Ipe's functionality between Ipe 5 and Ipe 6. René van Oostrum insisted that no self-respecting drawing program can do without style sheets and layers. Views allow you to incrementally build up a page in a PDF presentation.
I also revised the interface to ipelets (which used to be called "Iums" in the good old days when people still thought that "applets" were small apples)—it is now based on dynamically loaded libraries (a technology that was still somewhat poorly understood in the early nineties, at least by me).
And, of course, there was a Windows version of Ipe 6. Who would have thought that ten years earlier!
There were many releases of Ipe 6.0, all of them called "previews", because I never considered that I had reached a stable state. A number of experimental features were tried and either built into Ipe or discarded. Ipe 6 migrated from Qt 2 and Qt 3 to Qt 4, a somewhat painful process due to a number of annyoing Qt bugs that cost me a lot of time.
When in 2007 I discovered the fantastic Cairo library for rendering, I immediately decided to switch Ipe to use this: a small dedicated library with a nice API to do the rendering, instead of the buggy monster that was Qt. The Cairo API fit Ipe so well that I could write a Cairo painter for Ipe in an hour or so. Cairo supports Freetype directly, instead of Ipe having to render each glyph into a bitmap that is then blit onto the canvas.
I made the huge mistake of announcing on the Ipe discussion list that Ipe 6.0 preview 28 was the last version of Ipe 6, and that there would soon be a new version, Ipe 7. I should have known that this was impossible during a time where I advised several graduate students, taught several new courses, and went through the tenure process. I had to release several bugfix releases of Ipe 6 while really wanting to work on Ipe 7.
However, the delay left me with enough time to carefully think about another change I wanted to make: It would be nice if Ipe embedded a scripting language that could be used to write simple ipelets without compilation. I looked at Scheme/Guile, Python, and Lua, and finally decided for Lua: a small, elegant, stable language with a tiny footprint, easily embedded with a very nice C interface.
In 2009, I had my first sabbatical ever, which I spent in the group of Ulrik Brandes at the University of Konstanz. Here I finally had the time to work on Ipe 7, and I'm very grateful to Ulrik and all members of his group for the wonderful time I had in Konstanz. Next to the two big changes mentioned above, Ipe 7 introduced tiling patterns, gradients, clipping paths, transparency, user-definable arrows and marks, and SVG output.
I wanted to avoid Qt in Ipe 7 as it had caused me quite a bit of pain during the life of Ipe 6, but it was hard to find a good replacement that would allow Ipe to run on Linux, Windows, and Macs. During the Korean Workshop on Computational Geometry organized by Tetsuo Asano at Hakusan seminar house in June 2009, I discussed using Ipe on tablet PCs with Vida Dujmovic, Jit Bose, and Stefan Langerman. It is their fault that Ipe 7 comes with a tablet input tool, and finally Stefan and Sébastien Collette convinced me that there isn't really an alternative to Qt that has the same support for tablets and Macs. So Ipe 7 is still using Qt, but in a much more restricted way than before—it turned out that this part of Qt is quite stable and not prone to new bugs.
Even though I was only using a small part of Qt to create the UI, I still had to bundle the huge Qt library, ten times the size of Ipe itself, with Ipe's binary builds for Windows and OSX. This bothered me, since of course Windows and OSX already provide all the functionality for building UIs natively. After some experimentation with building simple UIs in Windows and GTK, I settled on a new architecture where the UI library would only be used in the Ipe program itself, in the Ipe canvas, and in the small Lua library ipeui that can be used to build dialogs from Lua (and which is entirely independent of Ipe). The native Windows UI was finished while I was visiting Emo Welzl in Zurich in early 2015, on a small Windows-tablet, and was first released in Ipe 7.1.7. During my sabbatical at Bayreuth University in 2016 Christian Knauer and his group provided me with a Mac in my office, providing both the opportunity and the time to learn Objective C and Cocoa. Ipe 7.2.1 was released in Bayreuth with the first native OSX UI.