Since I work as a systems/network admin, much of my day is spent either building new systems or doing maintenance on existing ones. Like most places, we rely on different types of monitoring software to keep us apprised of system performance.
One such system is Cacti, which we have implemented recently and really like. It uses Apache, PHP, MySQL and RRDTool, generates performance graphs for lots of networks and systems, and is so dead easy to use compared to our previous system using MRTG and RRDTool, that I don't know how we worked without it.
Consequently, I've spent a lot of time in the last two months with a web browser window open in Cacti, either passively monitoring a graph, or adding new ones. We're bringing on lots of new DSL customers so I have been adding new graphs on a daily basis.
What I've discovered is that Safari is a horribly written browser, and that it is totally unusable for the simple function of opening a page, displaying it, and letting me click from page to page inside Cacti.
The first problem I encountered was with Safari 1.2 and 1.3 under Mac OS X 10.3. Safari 10.2 would take forever to open a page with lots of data in Cacti. For example, when you add a new device to Cacti, it does an SNMP query and shows you all of the available interfaces it can graph. On our router that terminates DSL customers, this is several hundred interfaces. Safari 1.2 totally choked on this page. I'd get the spinning beach ball of death, and would eventually have to kill the browser entirely.
Safari 1.3 that came with OS X 10.3.9 was even worse. This data-intensive page would simply cause the browser to crash totally after about 30 seconds. Also Safari 1.3 would crash if I had more than two SSL pages open.
Safari 2.0 in Tiger doesn't have the crashing problems described above, but it still takes the browser somewhere on the order of 5 minutes to display the page with hundreds of interfaces. During that time the browser is unusable with the spinning beach ball.
The killer problem that I just noticed today is that Safari positively EATS memory. I was working in Cacti, and noticed that my system had really slowed to a crawl. The browser window took forever to update, and I had a lot of disk activity. I fired up top and WHOA! I only have 2M of free RAM and the system is swapping like mad.
Safari was taking up 216M of RSIZE memory in top. That's half of my system total! Again, this is a single web site. I have a single window open, at this time displaying a list of 25 graphs which I was editing one at a time.
216M of RAM! What on earth are you doing to me, Apple?
I started up Firefox. Worked for a while on the same page, same tasks, about the same amount of time. Opened more tabs. Watched top in the background, and Firefox slowly went from 32M after opening, to fluctuating between 36 and 40M depending on how many windows I had open.
I'm sorry, Apple, but your browser just sucks.
Does Safari suck for you too?
Categories
Web
Comments (7)
Read More Entries by Roger Weeks.
I've come to the conclusion that Safari such on sites with heavy Javascript.
It seems like any large site with extensive Javascript drops Safari to it's knees. Here's a test for you. Get out your stop watch and time how long it takes to load FedEx's shipping portal on Safari. Next, do the same test on IE for Windows. (You can pretty much use the time it takes to load on Safari to go out, buy a windows machine, install it, load IE and open the FedEx page before Safari is done... Perhaps that's slight hyperbole, but you get the poine.)
Safari sucks memory
and just plain sucks!
Powerbook + tossed though one of Apple's pretty store windows = happiness.
Oh, and until Apple makes its own brand router, it would be nice if the ONE I BOUGHT AT THE APPLE STORE, worked with my base station.
a bit harsh
I understand your annoyance. But I think it's a bit harsh to say that Safari "just sucks", and that it's not suitable for use as a browser at all, just because you've found one web application with which it doesn't work well.
It sounds like Cacti produces very large pages. Given that most pages on the internet aren't that large, maybe Apple have concentrated on making Safari good at loading all of those?
Still, you're quite right that Safari shouldn't be behaving the way you describe. Hopefully they'll take a look at the problem - but yelling "you suck" at them probably isn't the most constructive way to start, especially when lead Safari developer Dave Hyatt has been so informative, and open to feedback, on his Safari blog (http://weblogs.mozillazine.org/hyatt/).
Virtual memory, not Safari per se?
Cacti is javascript-intensive. PHP has nothing to do with javascript.
Virtual memory, not Safari per se?
It was on the third iteration of Safari, with different problems each time, that I finally lost it, and simply had to post the details.
I haven't been able to recreate the memory creep issues with other pages, but the Cacti display pages don't use JavaScript, just PHP.
The reason my system was using VM is that Safari was sucking all the available RAM. So I don't see this as a VM issue. If I had more RAM I would assume that Safari would continue sucking it all up.
Wrong subject line, sorry!
Subject line on above ought to have read: Individual pages will throw any browser...
No browser is perfect and either am I!
My apologies...
Virtual memory, not Safari per se?
I'm sure you can find individual pages that screws any one browser. Not necessarily big pages either! Most browsers are fine for the vast majority of pages, but specific things will throw any one of them for a loop.
Firefox for example has issues (JavaScript related, I believe) that mean that I never use it for certain pages and use Safari for these instead.
Browser issues are annoying; file a bug by all means, but I wouldn't have thought the fact that one specific website throws a specific browser is worth this space! ;-) JMHO. (Though I'm sure you feel happier to get it off your chest...! ;-) )
(You've ruled out buggy JavaScript or the like? "Runaway browser" issues sometimes prove to be bugs in the website code which different browsers have different tolerances for.)