cohen@santa_fe.Tops.Sun.COM (Michael Cohen) (02/16/90)
In article <1990Feb14.184128.738@intercon.com> amanda@mermaid.intercon.com (Amanda Walker) writes: >example, copying files via TOPS is optimized for LocalTalk, which means it ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >doesn't try very hard to push data across any faster. Public Folder does >a better job of pushing the bits across the wire I must object to this claim. The performance of TOPS vs. PublicFolder for copying files has nothing at all to do with whether TOPS is optimized for LocalTalk vs. EtherTalk (or IP or TokenRing, etc.). It has to do with the fact that TOPS is a file-sharing product, not a file transfer product. Hence TOPS uses a protocol that is optimized for file-sharing, namely a request-response transaction oriented protocol, such as ATP. PublicFolder on the other hand is a file-transfer product, so it uses a protocol which is optimized for file transfer, namely a stream-oriented protocol, i.e. ADSP. - Michael Cohen TOPS cohen@santa_fe.ebay.sun.com disclaimer: my views, not those of TOPS.
peirce@claris.com (Michael Peirce) (02/16/90)
In article <741@suntops.Tops.Sun.COM> cohen@santa_fe.tops.sun.com (Michael Cohen) writes: >In article <1990Feb14.184128.738@intercon.com> amanda@mermaid.intercon.com (Amanda Walker) writes: >>example, copying files via TOPS is optimized for LocalTalk, which means it > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >>doesn't try very hard to push data across any faster. Public Folder does >>a better job of pushing the bits across the wire > > I must object to this claim. The performance of TOPS vs. PublicFolder for >copying files has nothing at all to do with whether TOPS is optimized for >LocalTalk vs. EtherTalk (or IP or TokenRing, etc.). > It has to do with the fact that TOPS is a file-sharing product, not a file >transfer product. Hence TOPS uses a protocol that is optimized for file-sharing, >namely a request-response transaction oriented protocol, such as ATP. > PublicFolder on the other hand is a file-transfer product, so it uses a >protocol which is optimized for file transfer, namely a stream-oriented >protocol, i.e. ADSP. Ahem, Public Folder does not use ADSP. It uses ATP. I agree that TOPS, being a file server, and Public Folder, being a file transfer utility, are optimized for different operations. Public Folder tries to blast files across the network, nothing more. TOPS must keep track of lots of different things as files are opened, closed, read, etc. Of course, this is also why Public Folder takes up so much less memory that TOPS does... One other comment about transfer speeds. Another variable people haven't mentioned yet is disk speed at each end. I've noticed that this can make a big difference for file transfers - especially when using EtherTalk. Claris Corp. | Michael R. Peirce (author of Public Folder) -------------+-------------------------------------- | 5201 Patrick Henry Drive MS-C4 | Box 58168 | Santa Clara, CA 95051-8168 | (408) 987-7319 | AppleLink: peirce1 | Internet: peirce@claris.com | uucp: {ames,decwrl,apple,sun}!claris!peirce
cohen@santa_fe.Tops.Sun.COM (Michael Cohen) (02/16/90)
In article <10880@claris.com> peirce@claris.com (Michael Peirce) writes: >In article <741@suntops.Tops.Sun.COM> cohen@santa_fe.tops.sun.com (Michael Cohen) writes: >> PublicFolder [uses] ADSP. > >Ahem, Public Folder does not use ADSP. It uses ATP. Sorry about this mistake. I knew that Pete Helme's "Oscar" is based on ADSP, so I assumed that PublicFolder was as well. But I still think that the difference in performance between TOPS and PublicFolder [I'm accepting Amanda's word on this; I've never seen any test results] is because the products are designed to do different things. Claiming that it's because TOPS is optimized for LocalTalk is not relevant. On another note, I don't want to give the impression that I'm out to trash PublicFolder. Although I've never used the product myself, I've only heard very favorable reports. It's seems to be a cheap, efficient file- transfer utility. - Michael Cohen TOPS cohen@santa_fe.ebay.sun.com
amanda@mermaid.intercon.com (Amanda Walker) (02/16/90)
In article <741@suntops.Tops.Sun.COM>, cohen@santa_fe.Tops.Sun.COM (Michael Cohen) writes: > I must object to this claim. The performance of TOPS vs. PublicFolder for > copying files has nothing at all to do with whether TOPS is optimized for > LocalTalk vs. EtherTalk (or IP or TokenRing, etc.). I'm still skeptical, though no offense intended. The last version of TOPS I used was 2.something (2.1? maybe), and it was noticably slower than other file-sharing products (such as AppleShare, for example) when run over EtherTalk. Some of this may be due to the fact that the TOPS server is designed not to "take over" the Macintosh it is running on, but the fact that it isn't as good as other products at making use of the Ethernet bandwidth indicates to me that the basic design of the software is based on the characteristics of LocalTalk. > It has to do with the fact that TOPS is a file-sharing product, not a file > transfer product. Hence TOPS uses a protocol that is optimized for file-sharing, > namely a request-response transaction oriented protocol, such as ATP. > PublicFolder on the other hand is a file-transfer product, so it uses a > protocol which is optimized for file transfer, namely a stream-oriented > protocol, i.e. ADSP. I must have a bad memory, then, because I could swear the author of Pulic Folder has stated on the net that it uses plain old ATP... -- Amanda Walker InterCon Systems Corporation "Many of the truths we cling to depend greatly upon our own point of view." --Obi-Wan Kenobi in "Return of the Jedi"
amanda@mermaid.intercon.com (Amanda Walker) (02/17/90)
In article <742@suntops.Tops.Sun.COM>, cohen@santa_fe.Tops.Sun.COM (Michael Cohen) writes: > But I still think that the difference in performance between TOPS and > PublicFolder [I'm accepting Amanda's word on this; I've never seen any test > results] is because the products are designed to do different things. > > Claiming that it's because TOPS is optimized for LocalTalk is not relevant. Actually, I wasn't particularly trying to trash TOPS. It just came to mind as an example of software which does not give an increase in performance that is proportional to the increase in "raw" network bandwidth. Sorry if I stepped on your toes, Michael :-). Saying that the difference is because TOPS is a file-sharing product doesn't seem too relevant either, though, and saying it's because TOPS uses ATP is downright silly. ATP should give *better* performance (except in the presence of severe network delays), since ATP has a much smaller connection setup/takedown overhead, as well as a smaller amount of buffering and state that must be kept per session. But, whatever--all designs involve compromises of one sort or another. I don't really want to get in a flamewar over this; in the end, there are many factors that affect the "bang for the bps" that you get out of your network: disk speed & file system overhead, protocol scheduling, MultiFinder issues, network traffic loads, and so on. -- Amanda Walker InterCon Systems Corporation "Many of the truths we cling to depend greatly upon our own point of view." --Obi-Wan Kenobi in "Return of the Jedi"