OSNews

...lead balloon.

Or a Led Zeppelin?

OOo2.0 also supports 64K rows, whereas 1.1.x could only handle 32K rows.

Yes, I encountered a situation where this was a problem. Blech! Stupid massive CSV files.

I think the last version of the 1.1.x series I've used has been 1.1.3, and it was unusable with larger xls files. I have even tried to convert it to OO's native file format, but, to my surprise, that was even slower o.O. Unusable means opening a 3.6 MB (approximately) excel spreadsheet with about 10 worksheets in more than 10 minutes, whereas Excel opened it in cca 10 seconds. That was the point where I stared at it in disbelief, but it was that way all the time.

Ya know, OOo is written for mass appeal. I don't think anyone writing it was thinking "hey, what about the people who want to use spreadsheets for serious statistical work?" Well, someone probably did, and was probably told "we'll do that when we have time, these crazy users are harping us about releasing ."

If you work with big spreadsheets, use the best spreadsheet program available: Microsoft Excel. If you need an office suite to write articles, school papers, or any number of other small tasks: You can probably safely save yourself $300 and just use OpenOffice.
Of course, if you've already bought MS Office, you might as well use it until it's not supported or you see a feature you want somewhere else.

Stop demanding perfection people. You'll never get it from anyone. No one has time to sit around and write algorithm proofs for your Office suite, and they usually don't even have time to redesign the unexpectedly slow parts.

I'd hate to work on the OOo team, they get so much guff for things... Probably because people have put OOo up on a pedestal it doesn't belong on.

'Stop demanding perfection people' - I wasn't, if you were referring to me. I just pointed out that it really is slow sometimes (O.K., most of the time....). I mean, yeah, I own MS Office (academic edition) and I like some parts, hate other ones. E.g. in OO, it's really nice that you can do almost the same text formatting operations in Writer, Calc, Impress, Draw perhaps, whereas in Excel and Powerpoint, you can't do some useful stuff from Word. This is just my experience, I don't demand anything (although it would be nice if OO got faster, etc. ;]). I think if you don't need all the features Excel and Calc have, try gnumeric. I did, I love it for simpler tasks. It even opens some xls files faster than Excel itself . The best solution might be using all both Oo and MS Office, since they can actually nicely help each other sometimes.

'Probably because people have put OOo up on a pedestal it doesn't belong on.'
Dunno, seems to be the best open source suite in terms of features and stability (I don't think I remember it crash, could be wrong, though).

Duh, if only I had more time to learn C and C++ better ˘_˘.

If performance is a great issue one should use something like R instead of a spreadsheet. It is much faster than Excel at handling large amounts of data. We use R when analysing DNA chips.

R is yet another scripting language that the *NIX-heads mistake for an application (like TeX). You have to be crazy to think that it can replace OOo Calc or MS Excel. Nobody wants to learn yet another programming language just so they can generate some charts.

That's a really bad attitude you have there. You must understand some people, myself included, refuse to learn spreadsheets when they know it's just a weaker language than they have access to anyway.

People who can actually manipulate complex spreadsheets are perfectly capable of coding the same things, given a few weeks to learn to program. We're not even talking about difficult programming, we're talking about really simple math analysis stuff.

Spreadsheets have their place. But I'd agree that it's not in 19 page calculations..

A lot of people use spreadsheets as a database/tabulation program. Which is fine, for a small amount of stuff; but in the end it's mass overkill. The only reason it started is just because Excel is *that* good. If Excel had a history like Word, I don't think you'd see people using it nearly as often as they do: Nobody wants to wait one minute for their program to load so they can checkoff something on a todo list (but Word XP took that long to load ).

People who can actually manipulate complex spreadsheets are perfectly capable of coding the same things, given a few weeks to learn to program.

They may be capable of learning how to program, but why would they want to waste their time? They'd rather spend their time thinking about what to analyze, and what the results mean (doing actual work) rather than wasting time fixing syntax errors, and other work that the programmers should do.

I'm against the whole application as a scripting language design that a lot of *NIX software uses. To me it's like the developers are just passing the buck to the end users. They are saying: "We don't know how our end users are going to use our software, and we don't want to spend the time to find out. So we're just going to write some API, expose it in a scripting language, and let the users waste their time trying to figure out how to do something useful with it. We're all too busy reading Slashdot."

Fixing syntax errors? Only absolute beginners spend a real amount of time fixing syntax errors, well, beginners and c++ developers.

Like I said, really pathetic attitude. For some things, yea a spreadsheet is the way to go, but if you've got 3MB of data you're probably wasting your time on a spreadsheet.

The actual reasoning behind giving users a full out turing complete language is because it has the largest range of theoretically solvable problems: It's the most powerful. And, for thousands of years, man has solved math problems on paper with math equations: Not with graphical boxy representations of what the data might almost be contorted to sorta look like.

I think many people doing actual statistics work would appreciate a powerful language for expressing their equations, rather than a graphical expression. Maybe it's not for everyone, but I'd say that as a rule mathemeticians/statisticians are much more equation oriented than spacially oriented anyway.

But you're right. Not assuming how your end users will use your software is a facet of Unix design, and it's probably the biggest reason why Unix designs are so far superior to their "broken down wizard" counterparts. Not dumbing things down is important, it says "hi, I'm a developer, I'm writing this for users whose intelligence I respect." Instead of "hi, I'm from marketing, this will do everything for you!"

Not assuming how your end users will use your software is a facet of Unix design, and it's probably the biggest reason why Unix designs are so far superior to their "broken down wizard" counterparts.

You misspelled knowing and inferior. But knowing some *NIX programmers, I could totally see how they would rather assume how their software is being used rather than go out and find out. That requires * GASP * social interaction.

Thank you for ceding the argument to me by insulting nameless people in a generic and derogatory manner.

I was just about to say people who don't register a nick are stupid, but you beat me to the punch!

When did I say that R could replace spreadsheets? You are jumping to conclusion because I didn't make that statement. R is a language and it is very efficient at processing large data sets. Exactly as I said. R is not very relevant for most users. However it would be relevant for that guy with 3 days of 10hz data.

Do not comment without at least some knowledge (you could google a few secs...), R is a statistics software and a damn good one at that. Ever heard of S?

Z D and J, that is just stupid.

/Lemma

Lets Google about 'R'

Obviously you didn't: http://www.google.com/search?q=r

Err, no, I'm sorry, but you are the twit, sir (yes, I do mean that). Why don't you stop trying to insult other people, and start reading what other people write?

Any person, even you, should be able to type just "R" into google, and get back a result. That's exactly four keystrokes for me: Ctrl-K (Firefox shortcut for Google search), R, then enter. Not so hard was it?

And lo and behold, guess what the first reuslt is?

The R Project for Statistical Computing
R, also called GNU S, is a strongly functional language and environment to
statistically explore data sets, make many graphical displays of data from custom ...
www.r-project.org/ - 2k - Cached - Similar pages

cya,
Victor

>R is a statistics software

And why do you think the rest of the world would know about it? Talking about "just" R as if it is well known?

So I found very convenient this asnwer:
>Z D and J

I even laughed with this becasue I just thought about a similar answer when read about "R", the point is that you wanted to look cool talking about "R" so the rest of the world would google it -like thinking "R?, what's that? maybe a wonderful app I didn't know until today? I have been living all these years and R was just there, my life could have been a lot easyer" ...

well I could continue. So don't talk like that
>Z D and J, that is just stupid.
When it's you who started talking about "just R", I repeat R!, great name!

OK, I tried looking at the memory used by OO.o 2 and OfficeXP with the spreadsheets he gave loaded. And wow, it was 233MB to 43MB. Maybe all the extra time spent loading was in allocating memory.

On a side note, it seems that this test really favors the Athlon architecture over the P4, at least in OO.o. I only have MS Office on an older computer so I couldn't compare those times (it was still faster).

What's more, most of that memory gets paged out to disk, and stays there. While I can't measure it on Windows, I know on Linux that the amount of physical memory used is around 10MB. Calling it a memory hog doesn't make much sense therefore.

In any event, 30MB is not that bad these days, that's what Firefox uses in a typical session.

Also, it's notable that besides the file save and load benchmarks, the author gives no other example of where OpenOffice uses a lot of CPU. Concluding that the whole suite is a CPU hog based on loading and saving files in one application (and not typical files at that) is a bit rich.

And as far as Excel's memory usage goes, I think this screenshot (off my PC) is interesting.

http://www.bfeeney.uklinux.net/images/excel-mem.png

On the left, you see the Excel executable file, which is 9.60MB in size according to the details pane. When you start a program, the entire executable is loaded into read-only memory, so we would expect Excel's memory usage to be around 10MB. However, if you look to the right, you see the Excel memory usage, as measured by the Windows task-manager, is 1,500K in size.

That's rather odd ;-)

What these people fail to realize is that MSFt loads about a dozen extra libraries into memory at boot time just for Office. You can't measure the size that way.

What i want to see is someone loading Excel through crossover Office on a linux machine and then posting the memory usage. i bet they are a lot closer.

In the end though MSFT has had 12 revisions to make Excel run faster. open Office is basically on 2. and it's a whole lot faster and more complete than Excel was at that time. Hopefully this will kick start the programmers to clean up the mess a bit.

In the end though MSFT has had 12 revisions to make Excel run faster. open Office is basically on 2. and it's a whole lot faster and more complete than Excel was at that time. Hopefully this will kick start the programmers to clean up the mess a bit.

Not exactly true. Marco Borries founded StarDivision 21 years ago to develop an office suite. StarOffice went through over 15 years of commercial development as a proprietary office suite before Sun opened it up as OpenOffice.org on the basis of the 5.x release of StarOffice.

The bloat is historical and is part compounded with a history of mixed proprietary and open source development. Remember when Netscape open sourced Navigator, the Mozilla group had to give up on the old codebase and effectively start from scratch as it was so convoluted.

I used Star Office about 5 years ago, I can assure you it's nothing like OOo. Maybe they managed to use a large amount of code from it.

The problems here are:
Excel is the best program Microsoft ships, it's their flagship "we have great geniuses working for us" product.

OOo is largely in Java, you can't check the memory usage. Try, go ahead, you'll get a wrong answer. It'll be as accurate as threaded programs memory footprint on Linux (totally wrong).


Those create and load times sound bad, but I wouldn't worry too awful much. You should see what kspread did with that 3.6MB SXC file... My athlon with 1.25GB of RAM ran out of RAM trying...

Taking 5 times as long as excel to work with this is quite respectable... And, the XLS doesn't count, of course; we don't count binary dumps as save operations here in the "we like being able to open files with more than one program" world.


I can pretty much guarantee for you that Microsoft Word does not sit inside 9MB of memory. Maybe it only consumes 9MB per process, but that much code will never reside in 9MB of RAM....

OOo is largely in Java

No, the core code for Writer and Calc which is derived from StarOffice is in C++, Sun and the OOo people added some Java but the only part of OOo you require a JVM for is the new Database package. The OOo hackers seem to have added a lot of Python stuff and as I remember it a Python interpreter and most of the Python environment is bundled is the OOo package.

Taking 5 times as long as excel to work with this is quite respectable

Here I agree with you especially as if you have a spreadsheet file as big as 3.6M you ought to be using a more sophisticated application to handle the data.

"the only part of OOo you require a JVM for is the new Database package"

Incorrect. Java is required for: the media player; mail merge to email; Writer document wizards; accessibility tools; and much more. See http://en.wikipedia.org/wiki/OpenOffice.org#Java_controversy for more details.

OpenOffice.org is bulky and slow enough, without requiring an entire virtual machine, language and libraries just for some fundamental features. There is a lot to be done...

Like a custom script .

If you would like to guage the author of the article's opinion, please read his other one-sided pieces [of shit.]

I have no doubt that MS Office is faster in many ways considering how few layers it has compared to OpenOffice.Org.

But it didn't cost me $499.


Excel is FAR more advanced than Calc, but calc has seen significant improvement from 1.1.5 to 2.0 in terms of features and useability, specifically graphing.

though sort of off topic, I am a regular user of OO.o Impress for creating flowcharts and the difference between 1.1.5 and 2.0 is VERY LARGE and it is VERY easy to use now.

No, we are talking about the latest version of OO compared with the latest version of MS Office. The version numbers don't matter, these are the versions that directly compete against each other.

Besides, MS Office didn't go 1.1.1, 1.1.2, 1.1.3, etc.

OO is 5 years old, sure MS Office is alot older, but then many of those years of development have been rendered irrelevant by the change in technology. Does the development on Office from 1990 till 1994 really have much positive impact of Office 12? I highly doubt it.

- Kelson

Its not just a few fps. OpenOffice 3-4 times as much memory and CPU cycles.

Granted I have used Word and Excel much more than OpenOffice but I used OpenOffice exclusively for my MBA and I found Calc to be infuriating at times. I really wish I could use 1-2-3 style '/' to access common functions. Charting on Calc was horrible. The ability to create sections in Word is superior as automatically inserting tables of contents.

OLE is also much better with MSOffice.

No it doesn't.

Fact is, that if you calculate the combined resource usage of socalled system libs in windows and Office applications, the numbers get very different

However, the test made by the astroturfer George Ou (remember his last test?) shows that there are issues when converting from one format to another, especially in regard to wrong usage of the formats.

No sane person would do what George Ou did in these files.

I'd say that opening a file is a quite common task.

> A more interesting benchmark would be to measure the total time it would take for an average user to accomplish certain common tasks using different office suites.


OK, I'll do just that. I will often read my work emails at home. I log on to the work VPN, and download my emails (via POP3) to my local PC. When I click on a Word attachment, it takes over a minute before the OO word processor displays the Word document.

Before I switched to OO on that PC, I had used MS Word '97. That took less than 10 seconds before the document was displayed.


btw, Thunderbird (I know it isn't OO, but situation is applicable) has a similar problem with start-up time. Why it takes an email client nearly a minute to start-up is beyond me. On the same PC I use Calypso and Mulberry, both start up in under 10 seconds.

The start-up times for OO applications are so slow that I could not recommend their use in an environment where "time is money."

Which is just as useless, because your common tasks may be different than my common tasks.

Tho, creating and loading documents does seem to be a pretty common component of any task, as I can't see how you could end up with a document w/o creating or loading it first.

Instead of everyone jumping on the anti-microsoft bandwagon and attacking the benchmarks, maybe they are correct and the OO folks just need to fix their create & load times.

- Kelson

> Which is just as useless, because your common tasks may be different than my common tasks.


Useless to you, but not useless to me. However, I do agree with the gist of your comment, and its implication that a great piece of software is a critical balance of features.

The Open Office team (and the Thunderbird team also) have some work to do. That is very apparent to me.

Mr. Ou's file contains more than 1 sheet. It contains about 10 sheets.

10*12 = 120 seconds

You have the idea, but my sample was 16 sheets. That makes it even worse.

The guy was fooling himself by using cut-paste to only replicate 1 sheet. All he needed to do was take my SXC file and do a "save as" to create a native ODS file. He didn't even get the extention "ODS" right.

Ok, I downloaded your file, and tried it in my OOo.
Like I said earlier, I use version 1.1.5 and it took 22 seconds to open the file. CPU usuage was at 99% during the loading and for 3 seconds afterwards, then it dropped down to 23%.
I concur with you, the guy should have used native formats or open formats for both, as it just shows his results as skewed or that he is a shill.

I just did, and you're right. I assumed that if it had an OO.o file type it was already converted, but it looks like OO.o tries to keep the data as intact as possible, converting it each time.

Hold on... Your document only has 1 sheet. Not 16 like Mr. Ou's. If I multiply the new time out by 16, it still comes out faster, but only about 15-20%. And that may be due to rounding errors and the complexity of additional sheets.

Yup. You hit it. There is something to fix about the way people use Excel, and something to fix in regard to converting the formats.

George Ou's example file does not resemble typical use of Excel.

It's a limit of whatever type is used to store the row/column number. In this case, a short is 16 bytes, which comes out to 64000 rows (unsigned) or 32000 (signed). I wouldn't think it would be too difficult to change to an integer, but I bet a fair amount of code would need to be modified. And it could potentially skyrocket the memory requirements. I have no idea how often the row/column numbers are used, but if they are being used in any large arrays it could be a big difference.

haha

before complaining you should at least know that excel is not meant for such kind of usage. oh well...

"Why when ever somebody say something bad about a Opensource product, a trillian zealots come out of no where and start bashing this guy?

People have thier own opinion!"

Dude, he's misrepresenting facts to support an opinion. That's just playing dirty.

We'll see about it when the real facts get on the table. This article was pure BS from start to end.

However, it's true that MS Office is faster (since most of it is being started beforehand, during system startup).

In regard to memory usage it's difficult to say, since most people don't count in the shared libraries of MS Office, but do count in the shared libraries of OpenOffice. Just like George Ou did.

His tests are completely flawed.

The only thing we can say, is: There are certain issues i OOoCalc around wrong usage of spreadsheets in Excel in regard to converting these to native Calc format.

That's all we can use it to.

"However, it's true that MS Office is faster (since most of it is being started beforehand, during system startup)."

No. MS Office starts up far, FAR faster under WINE on Linux. It leaves native OpenOffice.org in the dust. That's with no preloading, no 'Microsoft Magic', no nothing. Just simply, sadly, better engineering and less inefficient, inelegant code.

OpenOffice.org is slower than MS Office for one reason: it's not coded as well. And yes, that's hard for some people to hear, and it does make you wonder just how powerful open source is when these problems come up (also see memory leaks and sluggishness in Firefox). But it's a fact that the sooner we realise, and the sooner we focus on it, the sooner we can make something faster, sleeker, more elegant than MS Office.

No. MS Office starts up far, FAR faster under WINE on Linux. It leaves native OpenOffice.org in the dust. That's with no preloading, no 'Microsoft Magic', no nothing. Just simply, sadly, better engineering and less inefficient, inelegant code.

There was a video on Channel 9 with Raymond Chen a while back. He talked about how to increase app start-up time. The gist of his argument was that, during start-up, the number of page faults has a huge influence on the speed. He said that you should profile your code, find the parts that are executed at start-up, and make sure that they are physically located next to each other in the compiled executable. When your app is loaded from the disk, caching will work better and it will dramatically improve the performance.

However, it's true that MS Office is faster (since most of it is being started beforehand, during system startup).

Do the Crossover Office guys have access to this too? When I load Microsoft Office in Crossover Office it still loads a lot quicker than OO.org.

I bet if this article said that OO.org was benchmarking faster than MS Office you wouldn't see a single comment saying that the test was flawed. It's typical OSS response. Say something in favour of OSS and we'll agree. Say something not in favour and we'll come up with every excuse in the book. Don't believe me, look at any comment thread when there's a bug in Internet Explorer. Now look at a comment thread when there's a bug in FireFox. You'll see the same people that say Microsoft can't code for IE, say that every piece of software has bugs when it's a bug in FireFox.

If this test showed OOo was faster in loading on Windows than MS Office, I would definitely like to see how the benchmark tests were done, since it doesn't fit with user experience at all.

So you're negative comments about typical OSS response are not only offensive, but obviously wrong. Of course there are some stupid zealots here and there, but forget about them (we could say a lot about all the windows zealots, and the mac zealots, the gnome zealots and the kde zealots, the beos zealots and all the other zealots if we wanted to), and look at the facts.

Wine is for an instance faster than XP in certain areas and slower than XP in other areas.

But isn't it funny, that benchmarks showing MS Office to be much faster and using much less memory always tend to use examples not likely to happen in reality? How come you can find so many flaws?

The sad part is that the test isn't that bad - George Ou just draws the wrong conclusions. And that's the real issue here.

The issues about bugs in IE vs. FF can be boiled down to the fact that in FF bugs get fixed must faster, and there are a lot fewer of them (including security holes according to Secunia). Many bugs in IE has been there for many years and are not likely to be fixed (at least not for a long while). The same goes for many other parts of Windows. Just look at scaling, or drag'n'drop. No fixes here since the first release of Windows4 (Windows95 and NT4).

But yes, you can of course find some who are blind in regard to flaws in open source software, just like there are people blind to flaws in proprietary software. See if you can navigate around them, even if it can be difficult.

You're wrong. The memory stats are from the amount of memory used by the process. Doesn't matter if it's a system DLL or app DLL, if the process uses that DLL, it counts against the memory usage for the process.

It doesn't change the fact that he doesn't count all processes in regard to MS Office apps

The attached link is unusable. It's only true in regard to mac-port.

The mac-port is known to be a lot worse than all other ports.

Let's see another test, this time comparing apples with apples, on the right platform (in this case the windows platform), instead of claiming that results for Mac-ports are true also for Windows ports.

Having used OOo 1.1, and OOo 1.9.xxxx, I'm pretty psyched about how much less crappy 2 is. We did give Microsoft a hard time for a bloated Office suite, remember? Office XP. But, it was bloated for a different reason: It had so many features you didn't care about you could hardly find the ones you wanted.

I'm not a big fan of OOo. But, I'd rather see people on a broken up bloated piece of FOSS than a bloated up broken piece of proprietary software. Why? Because 5 years ago Trident was what Trident is now. 5 years ago gecko barely worked, it's really nice now. 5 yeard ago khtml was pathetic, it's competitive today. I think in 5 years OOo will be vastly superior than it is today. The only reason Office has improved is because it's 60% of the company's income and they can't afford to lose it...

I suppose we don't yet know what happens to a FOSS project when it runs the market.

So, when are you starting to work on this project?

Huh, when you quit trying to sound smart maybe?

I think it might be more productive to just dump OOo and work on improving Koffice instead, this one was done right from the start. And guess what? When it will be ported to Qt 4, it'll use even less memory.

"so what do you expect for £200 and one of the best, richest acclaimed software developers in the world?"

So that's it? I remember when free software used to be about so much more -- the power of open source coding. Now, with bloated, buggy and security-hole-prone open source apps in the limelight, we're just back to price, it seems!

I agree with you, these kinds of tests are rather irrelevant. No sane user do things like this, if you
have this much data a relational database would be a much better choice.

However, OOo also target windows users and they are not used to having access to such tools. Until resently they couldn't just download postgresql and do some quick stored procedures.

To them, such functionality came at the price of a MS-SQL server license. Because of this, it's no wonder that they have strange user behavior, and naturally they try to transfer such behavior to OOo if they decide to switch office suit.

Learning to do it the right way will be much more costly to them than just continue using Microsoft products. So if OOo wants to compete it got to be faster even at this kind of extreme usage.

Damn right! We only want facts if they favor open source.

You can make anything look like a FACT and probably fairly easily design a benchmark that proves the opposite. No, I am not a zealot but if you haven't read anything from ZDNet's George Ou before perhaps you might want to get familiar with his style and see for yourself how "neutral" he is when it comes to anything open source.

you are simple and ignorant.