Hi Theo, it's hard to imagine a unix being written in anything other than C. But do you have feelings about how we should be writing code for application layers? Would you like to see less written in C, and more written in Ada, Scheme, Java or other languages that make leaks less likely. Have you played with Ada, do you think its type model is a general improvement over what's available in C?
If you're a songwriter then you don't have a business model. That's the way it should be. It's not the job of governments to run around handing business models to you just because you feel that you're entitled to it, which is all that copyright amounts to.
The economic case for copyright has never actually been made. To present it, you'd need to work out the economic activity that would occur in the world with it, and without it, and compare. There's lots of speculation along the lines of "how would blah be possible without copyright" but nothing approaching actual facts. Plenty of fantastic works were created before copyright, and many will be created after it. The trick is not to get sucked into the mental state of seeing what we have now that would be lost, but ignoring the potential that we would gain were it missing. I think the 'victims' of the removal of copyright would be only those rent seekers who up until now have done so well out of it - music industry middlemen, software monopolists, lawyers, copyright trolls. I've worked as a software developer for ten years, and copyright has not once been leveraged in a meaningful way in the numerous jobs I've been working in (nor has the assumption of it been related to the way we make money or the relationship with have with customers).
> Also I was beginning to think Slashdot was dying since > I hardly come here anymore, but with news like this I > feel mistaken.
Exactly. This is the sort of thing you could read about on the slashdot of old, then come to the comments section and exciting read stories by gurus who'd played around with interesting similar stuff.
I hardly ever come here any more, and it feels like this is the first interesting thing I've seen in eighteen months.
Hmm. Paypal isn't in that list. They've broken my account such that I can't use my existing account, can't sign up for a new one, can't get responses out of their technical support. They're the worst I've ever dealt with - Sony Online comes in second for me. They complete lost an order of mine and disappeared the payment.
I bought a black macbook this morning. Needed a powerfulish but portable dev server. Wanted black because the plastic probably won't stain with heavy use the way the white ones do. But I talked them down on price so as not to get ripped off.
It's now running Ubuntu. The new wireless card isn't supported. Setting up xmodmap has been painful and should be unnecessary (and it's still not as I'd like it) but even so - nothing else comes close.
> OS/2 died because no one high enough up the corporate command > structure lived or died by OS/2's success.
Exactly.
Further, for all the ridiculousness of the new vista interface, at least it's being done with the intention of impressing users. OS/2 never felt like it was being done with the users in mind. Maybe their bosses - but not the users themselves. There were annoying user interface issues with version 2 that still hadn't been fixed by version 4. New releases could come out supposedly with neat changes and those changes would be all these weird backend things that dovetailed nicely with existing IBM technology but which had absolutely no relevance to the enduser. Meanwhile you try to line up your icons and it still doesn't work nicely. Netscape had a memory leak that happened when a page changed and you had a dropdown open that would bring down the system. Didn't get fixed for years. Their advice on what to do for some DOS-origin games was "disable the sound". Gee, thanks. They didn't put particular effort into drivers.
I was completely into the OS/2 thing from 2.0 until 4.0 and by the end of the experience felt far more venom towards IBM than Microsoft. OS/2 lost because IBM couldn't tell its arse from itss elbow when it came to providing a reasonable user experience.
I would ask "Do you acknowledge acknowledge your sins and seek repentance in the name of our Lord Jesus Christ?" and then I would follow the response with "Your sins are forgiven my child, in nomine patris et filii et spiritus sancti amen."
I've got a Sony Clie and I use it occasionally for taking photos and somewhat regularly as my encrypted password store. It's good for those things. Those things are useless as organisers though (having said that - when I get snowed under I organise my life using my own text document format that I roll over each morning and edit in vim so I'm possibly not typical).
Thanks for your response. Where you're headed with the jabber idea is exactly the sort of think I'm thinking of - potential for a better world, etc. One thing I've wanted to get around to building with jabber for a while is some sort of jabber bot that sits there and takes command-line arguments. If you want to have a braindump about this stuff email me, craig@ the parent domain for my URL (leave off the 'stable').
> What exactly can't IMAP do that is IMAP (the protocol's) fault?
As I'm not familiar with IMAP I don't know what's the protocol's fault and what's the client's fault so I'm open to either, but the fact that I have yet to encounter an IMAP client that doesn't have a horrible user experience in ways that I (a not-naive analyst/programmer) can't find good fault for (the best I know of are thunderbird and apple mail and both are quite flawed) I'm open to the idea that there's something about IMAP that doesn't lend itself towards effective implementations.
> Those requests are about as useful as asking for them to make > the internet go faster."
It's not at all. Why can't we have a situation where the buffering between the client and the server is done in the background and doesn't interfere with the user experience? Sure you have to wait to get your mail, but thunderbird becomes brittle to use when it is latency-affected in ways that it should not.
A couple of respondants to my email have taken an approach of "Oh, IMAP doesn't support that." My response to that is - so screw IMAP. Why isn't there a mainstream attempt to create something that actually fulfils the requirements let down by IMAP? Perhaps because web mail is almost good enough.
> In what way is it "monolithic"?
To start with, the viewer and the composer are part of the same binary. Classic monolithic design. Why are they done in the same process? When you use IMAP functionality some things will freeze or fail to work properly when the client is talking to the server and busy things in the client stop working properly. Again, classic side-effect of monolithic implementation.
> It does the very basic requirements needed by an email client
Really. Takes ages to start up; contains spam controls; message filtering; HTML email editing... That's hardly basic functionality for an email client, unless you're conditioned into only conceiving of email clients in the model that we currently think about them or where you define basic by "that's what Outlook Express does".
I think if you sat down with a blank sheet and thought about what you really want out of email rather than what you've become used to accepting you'd think of all sorts of requirements which would seem quite reasonable and achievable that you hadn't previously thought of.
My blog is down at the moment because I'm working out some DNS issues with my secondary after a migration, otherwise I'd point you to an article pointing why Word is an absurd model for a word processor. Yet the mainstream swallow it because that's what's available - many of them do so under protest. How many times have you had to fix a computer where the owner has been malware affected and due to them using POP they lose all their email due to the error. Or due to a hard disk crash. Typical contemporary email solutions are horrible.
Yeah - you mention inadequacies of IMAP a few times. Which makes me wonder - why is there constant effort going into IMAP mail clients yet no effort to create a protocol that fixes up all the problems with IMAP.
Yeah - gmail is not perfect for lots of little reasons. There's not a way to keep mail on your laptop so that you can refer to it like when you forget the address of something that's in an email - I do this a lot.
I watched the demo. While this looks interesting, it's even more monolithic than the stuff I'm complaining about. It's not a native client tool, either. Web applications are cool for application purposes but... [I was going to say here that my gut feeling was that we needed a desktop-end client for email but wanted to give a 'real' reason and couldn't find one..]
Hmm. It would be possible to have fetchmail ticking away in case you ever wanted to check email offline, but then to use a web-based email system the rest of the time. That might work and wouldn't require any special new technologies.
That still doesn't cover addressing though. Although an intelligently designed ajax-friendly webmail client could fix that.
OK - what about the situation where you click on a mailto: link in your webbrowser? That wouldn't be covered. But it might not take a lot of work from the mozilla team to put functionality to support that into their browser. Or maybe it could be done via a plugin?
Thanks for the link - I don't think that's the answer to my issues but it has set me on a promising new line of thought.
In the last few years the browser platform has matured and after a long period of it being awkful, I'm content with the current state of things. But I feel that email has not improved at all over the last ten years. The only major change has been the rise of spam - a step backwards.
Some of the comments below will link to my lack of skills in areas of system administration and I encourage replies to those issues as much as any other feedback. Better yet - write a howtoforge article describing how to set such a system up under debian stable:)
My needs for an email system are: - data should be stored on the server (centralised backup, provision for web mail when you need it, ability to have an administrator control it, access from multiple hosts) - server-side spam filtering which can also take easily feedback from the client on what proved to not be spam, or what was and was missed. - server-side addressbook - should deal only with plain text - non plain text should be flattened to plain text. It would be nice to automatically bounce office files with a message to tell the person to send stuff as PDF or plain text. - effective searching - very responsive client for reading mail - very responsive client for writing mail - effective communication between client and server that doesn't require the user to wait
I don't really see how thunderbird's design lends itself to fitting into an infrastructure that meets those requirements.
Perhaps my biggest problem with Thunderbird and all mail clients that I've encountered is that IMAP proves to be inadequate. Communicating with an email server over IMAP makes for a klunky experience (*particularly* over a latent connection), and it shouldn't need to be this way. Perhaps IMAP is a bad fit for the task.
Time and time again we see people trying to build a 'Microsoft Word killer' without them ever stopping to think about whether a monolithic word processor is even a good idea (I suggest that it's not). Similarly, Thunderbird strikes me as a really good attempt at producing a product idea that is fundamentally flawed. We should be working to phase out monolithic email clients.
Surely all that should be required of a good client is this: - Keep the client's disk archive of the mailbox synchronised with the server so that searching is easy, and do so inobtrusively (all the IMAP clients I've used are quite obtrusive and brittle as the number of possible connections rises), but reflect changes to the client back on the server (I don't think fetchmail does this) - Composer that has access to the server's addressbook and sent folder and has a spellchecker - Email viewer
I must have really lucked out. Usually with linux and me, everything that can go wrong does go wrong. However, I've upgraded my desktop workstation and my development server (both running ubuntu, very different setups though) and both have been seamless.
This perspective makes absolutely no sense to me. I built a dual P2 between christmas and new years in 1998. It's still functioning as a server for me. Rock-solid. Though the course of its life it's had substantial periods running BeOS and Windows NT and now it's running Debian. No problems. I serve real applications off it.
I bought a PPC mac laptop early last year and it was slow when I got it. I knew that when I bought it and got it for webobjects and that's worked out well - but it's pretty obvious that it's not going to have fantastic life in it. I don't know how you were running a powermac 9500 and happy with it - I've had a powermac 9600/200 circa 1996 (and bloody expensive when new) sitting behind my door being useless for years now. True - it could be playing server but I never bothered because it weighs a tonne, it's hard to get more memory for, however good ppc/linux is I doubt it'll be as good as ppc/x86 and I don't trust the power management as much as on a newer intel machine.
I don't think Apple can compete bangs for bucks on what you can put together yourself, and that is a viable comparisson, because it's not possible to put your own Apple hardware together (or if it is I don't know about it).
> If I were the premier of China, I'd make a secret deal with SK to put a > military sqeeeze on the place, since NK would probably be overwhelmed by > a Chinese invasion.
I can't remmeber where I read this, but understand the main concern of all neighbours to be a complete collapse of the current system. If that happens there will be fifteen or so million refugees of nearly zero economic value suddenly sitting on the doorstep of two of the world's most significant economies. China doesn't want a North Korean collapse.
Even before nuclear weapons were a factor military intervention wasn't really an option as the north have artillery trained on Seol which is the second largest metro area in the world - http://en.wikipedia.org/wiki/List_of_metropolitan_ areas_by_population. And South Korea is economically significant (http://en.wikipedia.org/wiki/Republic_of_Korea) - so letting that happen is not viable to anyone.
Something I find infuriating about excel is that they've done something to break the way that the alt+tab ordering works. In the versions of the product I've used there's always a 'ghost' application sitting in the task list and it doesn't restack properly meaning you subconciously switch following the convention and then end up in the wrong place. Very very annoying.
I have, and have used cayenne in a couple of substantial apps. But that's not quite what I want. Or at least - I want that, but then I want another facade in between it. So I end up with
database business-logic-engine view framework
In the scenario I describe, the engine would look to the view framework like a database but once you 'committed' that would trigger a cycle in the engine. If the commit failed, it would throw a jdbc exception.
It's a new take on RPC. It might be horrible - I haven't completely got my head around the idea. But it could be nice because it would mean you could use existing tools with jdbc support to access your engine (through simple binding, that sort of thing).
> I haven't met a person who has *used* OS X for any length of time and not loved it
[raises hand]
I'm here.:) It has no native support for workspaces (and none of the avaialble solutions are elegant); aqua is bloaty; finder is awful and buggy as well; under mac os some unix tasks such as setting up groups are far more painful than under linux/solaris/bsd (I think you have to use this nicl utility that is quite a step removed from the 'everything is a file or configured through one' philosophy); there are long-term problems with terminal support under console mode; the last time I looked the package management systems were less mature than for other platforms; the hotkey support is inferior to Windows (or to gnome when it's not broken as it is in ubuntu 6.06); there is no responsive GUI mail tool that does IMAP properly (thunderbird is quite unresponsive); there's no ability to support meta+tab to switch between windows as in other popular operating systems (and none of the third party utilities I've tried work properly). I have an ibook for WebObjects, AudioHijack, Indesign, and the painless-unix-and-wireless-and-media-codecs-on-a-l aptop factor but I far prefer gnome and a linux||bsd||solaris command-line.
Slashdot Mirror
Hi Theo, it's hard to imagine a unix being written in anything other than C. But do you have feelings about how we should be writing code for application layers? Would you like to see less written in C, and more written in Ada, Scheme, Java or other languages that make leaks less likely. Have you played with Ada, do you think its type model is a general improvement over what's available in C?
If you're a songwriter then you don't have a business model. That's the way it should be. It's not the job of governments to run around handing business models to you just because you feel that you're entitled to it, which is all that copyright amounts to.
The economic case for copyright has never actually been made. To present it, you'd need to work out the economic activity that would occur in the world with it, and without it, and compare. There's lots of speculation along the lines of "how would blah be possible without copyright" but nothing approaching actual facts. Plenty of fantastic works were created before copyright, and many will be created after it. The trick is not to get sucked into the mental state of seeing what we have now that would be lost, but ignoring the potential that we would gain were it missing. I think the 'victims' of the removal of copyright would be only those rent seekers who up until now have done so well out of it - music industry middlemen, software monopolists, lawyers, copyright trolls. I've worked as a software developer for ten years, and copyright has not once been leveraged in a meaningful way in the numerous jobs I've been working in (nor has the assumption of it been related to the way we make money or the relationship with have with customers).
> Also I was beginning to think Slashdot was dying since
> I hardly come here anymore, but with news like this I
> feel mistaken.
Exactly. This is the sort of thing you could read about on the slashdot of old, then come to the comments section and exciting read stories by gurus who'd played around with interesting similar stuff.
I hardly ever come here any more, and it feels like this is the first interesting thing I've seen in eighteen months.
> Deal With
Thanks for pointing that out.
Hmm. Paypal isn't in that list. They've broken my account such that I can't use my existing account, can't sign up for a new one, can't get responses out of their technical support. They're the worst I've ever dealt with - Sony Online comes in second for me. They complete lost an order of mine and disappeared the payment.
I bought a black macbook this morning. Needed a powerfulish but portable dev server. Wanted black because the plastic probably won't stain with heavy use the way the white ones do. But I talked them down on price so as not to get ripped off.
It's now running Ubuntu. The new wireless card isn't supported. Setting up xmodmap has been painful and should be unnecessary (and it's still not as I'd like it) but even so - nothing else comes close.
> OS/2 died because no one high enough up the corporate command
> structure lived or died by OS/2's success.
Exactly.
Further, for all the ridiculousness of the new vista interface, at least it's being done with the intention of impressing users. OS/2 never felt like it was being done with the users in mind. Maybe their bosses - but not the users themselves. There were annoying user interface issues with version 2 that still hadn't been fixed by version 4. New releases could come out supposedly with neat changes and those changes would be all these weird backend things that dovetailed nicely with existing IBM technology but which had absolutely no relevance to the enduser. Meanwhile you try to line up your icons and it still doesn't work nicely. Netscape had a memory leak that happened when a page changed and you had a dropdown open that would bring down the system. Didn't get fixed for years. Their advice on what to do for some DOS-origin games was "disable the sound". Gee, thanks. They didn't put particular effort into drivers.
I was completely into the OS/2 thing from 2.0 until 4.0 and by the end of the experience felt far more venom towards IBM than Microsoft. OS/2 lost because IBM couldn't tell its arse from itss elbow when it came to providing a reasonable user experience.
I would ask "Do you acknowledge acknowledge your sins and seek repentance in the name of our Lord Jesus Christ?" and then I would follow the response with "Your sins are forgiven my child, in nomine patris et filii et spiritus sancti amen."
I've got a Sony Clie and I use it occasionally for taking photos and somewhat regularly as my encrypted password store. It's good for those things. Those things are useless as organisers though (having said that - when I get snowed under I organise my life using my own text document format that I roll over each morning and edit in vim so I'm possibly not typical).
Thanks for your response. Where you're headed with the jabber idea is exactly the sort of think I'm thinking of - potential for a better world, etc. One thing I've wanted to get around to building with jabber for a while is some sort of jabber bot that sits there and takes command-line arguments. If you want to have a braindump about this stuff email me, craig@ the parent domain for my URL (leave off the 'stable').
> What exactly can't IMAP do that is IMAP (the protocol's) fault?
As I'm not familiar with IMAP I don't know what's the protocol's fault and what's the client's fault so I'm open to either, but the fact that I have yet to encounter an IMAP client that doesn't have a horrible user experience in ways that I (a not-naive analyst/programmer) can't find good fault for (the best I know of are thunderbird and apple mail and both are quite flawed) I'm open to the idea that there's something about IMAP that doesn't lend itself towards effective implementations.
> Those requests are about as useful as asking for them to make
> the internet go faster."
It's not at all. Why can't we have a situation where the buffering between the client and the server is done in the background and doesn't interfere with the user experience? Sure you have to wait to get your mail, but thunderbird becomes brittle to use when it is latency-affected in ways that it should not.
A couple of respondants to my email have taken an approach of "Oh, IMAP doesn't support that." My response to that is - so screw IMAP. Why isn't there a mainstream attempt to create something that actually fulfils the requirements let down by IMAP? Perhaps because web mail is almost good enough.
> In what way is it "monolithic"?
To start with, the viewer and the composer are part of the same binary. Classic monolithic design. Why are they done in the same process? When you use IMAP functionality some things will freeze or fail to work properly when the client is talking to the server and busy things in the client stop working properly. Again, classic side-effect of monolithic implementation.
> It does the very basic requirements needed by an email client
Really. Takes ages to start up; contains spam controls; message filtering; HTML email editing... That's hardly basic functionality for an email client, unless you're conditioned into only conceiving of email clients in the model that we currently think about them or where you define basic by "that's what Outlook Express does".
I think if you sat down with a blank sheet and thought about what you really want out of email rather than what you've become used to accepting you'd think of all sorts of requirements which would seem quite reasonable and achievable that you hadn't previously thought of.
My blog is down at the moment because I'm working out some DNS issues with my secondary after a migration, otherwise I'd point you to an article pointing why Word is an absurd model for a word processor. Yet the mainstream swallow it because that's what's available - many of them do so under protest. How many times have you had to fix a computer where the owner has been malware affected and due to them using POP they lose all their email due to the error. Or due to a hard disk crash. Typical contemporary email solutions are horrible.
Yeah - you mention inadequacies of IMAP a few times. Which makes me wonder - why is there constant effort going into IMAP mail clients yet no effort to create a protocol that fixes up all the problems with IMAP.
Yeah - gmail is not perfect for lots of little reasons. There's not a way to keep mail on your laptop so that you can refer to it like when you forget the address of something that's in an email - I do this a lot.
I watched the demo. While this looks interesting, it's even more monolithic than the stuff I'm complaining about. It's not a native client tool, either. Web applications are cool for application purposes but ... [I was going to say here that my gut feeling was that we needed a desktop-end client for email but wanted to give a 'real' reason and couldn't find one..]
Hmm. It would be possible to have fetchmail ticking away in case you ever wanted to check email offline, but then to use a web-based email system the rest of the time. That might work and wouldn't require any special new technologies.
That still doesn't cover addressing though. Although an intelligently designed ajax-friendly webmail client could fix that.
OK - what about the situation where you click on a mailto: link in your webbrowser? That wouldn't be covered. But it might not take a lot of work from the mozilla team to put functionality to support that into their browser. Or maybe it could be done via a plugin?
Thanks for the link - I don't think that's the answer to my issues but it has set me on a promising new line of thought.
In the last few years the browser platform has matured and after a long period of it being awkful, I'm content with the current state of things. But I feel that email has not improved at all over the last ten years. The only major change has been the rise of spam - a step backwards.
:)
Some of the comments below will link to my lack of skills in areas of system administration and I encourage replies to those issues as much as any other feedback. Better yet - write a howtoforge article describing how to set such a system up under debian stable
My needs for an email system are:
- data should be stored on the server (centralised backup, provision for web mail when you need it, ability to have an administrator control it, access from multiple hosts)
- server-side spam filtering which can also take easily feedback from the client on what proved to not be spam, or what was and was missed.
- server-side addressbook
- should deal only with plain text - non plain text should be flattened to plain text. It would be nice to automatically bounce office files with a message to tell the person to send stuff as PDF or plain text.
- effective searching
- very responsive client for reading mail
- very responsive client for writing mail
- effective communication between client and server that doesn't require the user to wait
I don't really see how thunderbird's design lends itself to fitting into an infrastructure that meets those requirements.
Perhaps my biggest problem with Thunderbird and all mail clients that I've encountered is that IMAP proves to be inadequate. Communicating with an email server over IMAP makes for a klunky experience (*particularly* over a latent connection), and it shouldn't need to be this way. Perhaps IMAP is a bad fit for the task.
Time and time again we see people trying to build a 'Microsoft Word killer' without them ever stopping to think about whether a monolithic word processor is even a good idea (I suggest that it's not). Similarly, Thunderbird strikes me as a really good attempt at producing a product idea that is fundamentally flawed. We should be working to phase out monolithic email clients.
Surely all that should be required of a good client is this:
- Keep the client's disk archive of the mailbox synchronised with the server so that searching is easy, and do so inobtrusively (all the IMAP clients I've used are quite obtrusive and brittle as the number of possible connections rises), but reflect changes to the client back on the server (I don't think fetchmail does this)
- Composer that has access to the server's addressbook and sent folder and has a spellchecker
- Email viewer
I must have really lucked out. Usually with linux and me, everything that can go wrong does go wrong. However, I've upgraded my desktop workstation and my development server (both running ubuntu, very different setups though) and both have been seamless.
This perspective makes absolutely no sense to me. I built a dual P2 between christmas and new years in 1998. It's still functioning as a server for me. Rock-solid. Though the course of its life it's had substantial periods running BeOS and Windows NT and now it's running Debian. No problems. I serve real applications off it.
I bought a PPC mac laptop early last year and it was slow when I got it. I knew that when I bought it and got it for webobjects and that's worked out well - but it's pretty obvious that it's not going to have fantastic life in it. I don't know how you were running a powermac 9500 and happy with it - I've had a powermac 9600/200 circa 1996 (and bloody expensive when new) sitting behind my door being useless for years now. True - it could be playing server but I never bothered because it weighs a tonne, it's hard to get more memory for, however good ppc/linux is I doubt it'll be as good as ppc/x86 and I don't trust the power management as much as on a newer intel machine.
I don't think Apple can compete bangs for bucks on what you can put together yourself, and that is a viable comparisson, because it's not possible to put your own Apple hardware together (or if it is I don't know about it).
> If I were the premier of China, I'd make a secret deal with SK to put a
_ areas_by_population. And South Korea is economically significant (http://en.wikipedia.org/wiki/Republic_of_Korea) - so letting that happen is not viable to anyone.
> military sqeeeze on the place, since NK would probably be overwhelmed by
> a Chinese invasion.
I can't remmeber where I read this, but understand the main concern of all neighbours to be a complete collapse of the current system. If that happens there will be fifteen or so million refugees of nearly zero economic value suddenly sitting on the doorstep of two of the world's most significant economies. China doesn't want a North Korean collapse.
Even before nuclear weapons were a factor military intervention wasn't really an option as the north have artillery trained on Seol which is the second largest metro area in the world - http://en.wikipedia.org/wiki/List_of_metropolitan
Michael, are you able to supply more details about this situation - articles, etc?
> Absolutely -- that and Excel.
Something I find infuriating about excel is that they've done something to break the way that the alt+tab ordering works. In the versions of the product I've used there's always a 'ghost' application sitting in the task list and it doesn't restack properly meaning you subconciously switch following the convention and then end up in the wrong place. Very very annoying.
How long until Alan de Botton is producing three-part documentaries on 'pastiche' user interfaces? :)
I have, and have used cayenne in a couple of substantial apps. But that's not quite what I want. Or at least - I want that, but then I want another facade in between it. So I end up with
database business-logic-engine view framework
In the scenario I describe, the engine would look to the view framework like a database but once you 'committed' that would trigger a cycle in the engine. If the commit failed, it would throw a jdbc exception.
It's a new take on RPC. It might be horrible - I haven't completely got my head around the idea. But it could be nice because it would mean you could use existing tools with jdbc support to access your engine (through simple binding, that sort of thing).
> I haven't met a person who has *used* OS X for any length of time and not loved it
:) It has no native support for workspaces (and none of the avaialble solutions are elegant); aqua is bloaty; finder is awful and buggy as well; under mac os some unix tasks such as setting up groups are far more painful than under linux/solaris/bsd (I think you have to use this nicl utility that is quite a step removed from the 'everything is a file or configured through one' philosophy); there are long-term problems with terminal support under console mode; the last time I looked the package management systems were less mature than for other platforms; the hotkey support is inferior to Windows (or to gnome when it's not broken as it is in ubuntu 6.06); there is no responsive GUI mail tool that does IMAP properly (thunderbird is quite unresponsive); there's no ability to support meta+tab to switch between windows as in other popular operating systems (and none of the third party utilities I've tried work properly). I have an ibook for WebObjects, AudioHijack, Indesign, and the painless-unix-and-wireless-and-media-codecs-on-a-l aptop factor but I far prefer gnome and a linux||bsd||solaris command-line.
[raises hand]
I'm here.