In last week's shootout, where I matched Microsoft Excel against OpenOffice.org Calc, OpenOffice.org came out on the short end of the stick in my benchmark tests. The tests not only showed Excel to have superior memory-handling efficiency, but superior file creation and load speeds in both the XML and XLS binary format. Beyond my expectations, the blog was flooded with comments and went almost as high as 400 responses. Some readers flamed me for screwing up the test even though they didn't read the test method and didn't even try to repeat the test themselves. Other readers begrudgingly confirmed my results that OpenOffice.org was indeed slow but defended OpenOffice.org as a free and open standard alternative to Microsoft. One such reader is Frenchman Frédéric Jean, who wrote this thoughtful letter to me:
Hi, My mother tongue is French, so please excuse my spelling and grammar.
You seem to be on a crusade recently to demonstrate how XML is bloated and how OOo is slow to open/close files. You have published benchmark figures based on a test file. I don't dispute your data or your conclusions.
I am a daily user of OpenOffice, and also a long-time daily user of MS Office user. I have extensively used VBA on numerous occasions. I am both an "Office power-user" and a software developer. I am talking from experience here.
In my experience, MS Office starts quicker and loads files quicker. As for saves, the difference is not as noticeable (or I didn't notice it) but I assume it is quicker too. Great isn't it? Well, guess what, it does not matter in most cases.
Sure, you may find scenarios in which MSO is 10e32 times faster at doing X, but if X is a scenario only a tiny minority of users will ever encounter, it is a pointless exercise. If "much slower" is still quick enough in user time, it is also pointless. Frankly, you look like a mainframe guy saying mainframes are really fast at Word processing, that they can load and save documents much quicker than a PC (let us ignore the fact that mainstream word processing software may not exist on mainframes). The claim is true, but pointless. Mainframe makers still make truckloads of money selling their hardware to people with a real need for it, but they are the first to realize that the value they bring to the table can only be appreciated by niche (in numbers, but not in dollars) customers.
The world is full of examples like that : a (comparatively) expensive product is better at some specific task than another, cheaper one. Yet, the cheaper one is used by more people because it is OK for the average case and, well, it's cheaper, which is a real long-term advantage no matter how you slice it. I bet IBM used the TCO argument with their clients 15 years ago to counter the Windows momentum. On a 20-year horizon, Office becoming that kind of a product is a real possibility.
Not a certainty, but a possibility. Market leaders, when they are toppled, always fall prey to products once seen as poor replacements.
Now, for the meaty part. I have developed an elaborate OOo spreadsheet for the small company where I work. It is used for quoting the manufacturing of complex products with many, many production steps. It is not a full-blown "application", everything is done through formulas. There just are a lot of them, with lots of searches through tables. It takes a while to save and load (counted in seconds, not minutes). Granted, it would be nice if it saved and loaded quicker. The file is just about 55kb, which is nice since there will be thousands of these files in a few years. A database-based solution would be better (from a computer science perspective), but for this small company the investment does not make sense and they need my time for other, equally important things.
I wanted to demo the system to a Microsoft Office user. I saved it to XLS format. 4500kb. 4.5Mb. That is 82 times bigger. I thought to. myself : "No way! OOo must be doing something stupid here". So I opened the file in Excel, stripped all the formatting, and got down to about 3Mb (as saved by Excel). I then wrote a VBA routine that "rebuilds" a workbook by copying all the sheets, re-creating all the formulas (as if they were re-typed by hand) and patching the named ranges to correct the 32000 rows limitation in OOo. With formatting preserved, it did not go far below 4Mb.
Sure, the Excel version loads faster. About 5-10 times faster (I did not measure it). But the difference in file size is the deciding factor here. You certainly know about the space-time compromise. Two orders of magnitude faster, two orders of magnitude bigger. Wow, how unexpected. Excel files are usually bigger than OOo's, but not 98 times bigger. I also guess that OOo loads are usually slower than Excel's but not 98 times slower. One order of magnitude for both speed and size is closer to my experience.
Choosing extreme examples is only good at creating controversy. I guess the "creating heat" part of your mission as a journalist is as least as important as the "emitting light" part...
My personal example is not better than yours. It just shows that the criteria used to decide what software system is the best in a given case are not limited to start times, load times and save times. So while I agree with your conclusions about speed for these specific operations, I would and do touch OOo with poles much shorter than 10 ft. All of our internal documents are OOo-based, and it all works very well. I am a primary user of these documents, so I eat my own dog food. For the sake of the company, I don't care if development on OOo stops tomorrow. The current version fits our bill. Good enough mixed with free is a proposal hard to refuse. The rest is all bonus to us. I have several managers, all Excel die-hards, listening to me with an open ear. Once you remove "getting MSO for free" from their equation, Microsoft's software begins to look much less interesting. "Lock-in" is a concept they react to. OOo for internal documents, PDF export to send files outside, and a few copies of MSO is a compromise they are prepared to live with.
I want to thank Frédéric for his thoughtful letter, so I will say this in response: You are pretty much admitting that OpenOffice.org is significantly slower than Microsoft Office but you argue that it's "good enough." While I agree with you partially on that point, business users typically want the best and they don't mind paying for it. It's kind of like saying a motel can save you money, but I prefer staying in a nice business class hotel. Retraining office staff on a slower and less capable Office suite may not be worth the time and trouble. I do respect your opinion and OpenOffice.org does offer a nice free alternative. Ultimately it's up to the end user and business to determine if Microsoft Office has enough ROI to justify its price tag.